• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

The all DHT SET Headphone Amp

regal

Member
2004-01-04 8:41 pm
MD
For those of you who are unfamiliar with headphones this thread may seem a little over the top. But on an engineer's salary there aren't any speaker's or "listening" room's that can compete to headphones. If you are into older recordings or raw soundboards many were mixed on the fly with some talented pioneers wearing headphones. Many recordings were made in the 60's with ambisonic techniques before the word was invented.

Now designing an amp for headphones is not as easy as it seems, well its easy for Sennheisers but hard for about anything else of quality. Lets look at the extremes for dynamic headphones (I am not including AKG1ks which are just speakers on a headband.)

With headphones distortion at dynamic peaks can be much more "visible" to the ear, its easy to design a tube headamp tube amp gets messy in complex passages. When you set your goals for a headphone amp you sort of have to take manufacturer's specifications with a "grain of salt." IOW's the specs are made by the marketing department. The closest spec to reality is generally the maximum power handling spec, cause if they get that wrong there is a warranty issue.

So when I design a head-amp for myself (the only person I care about :D, I plug it in to a reference amp and turn up the volume to a healthy comfortable level playing a dynamic recording. Then I leave the volume knob in the same spot and play a 0 dbs 400 hz test tone and measure the ac voltage across the coils with my fluke.

The jist of all this comes down to we can set our design goals with these hard to drive but wonderful to listen to headphones and 2 "objectives" :

1. AKG 600 ohm sextettes - my test show they need an amp that can deliver 9 Vrms, thats only 9*9/600= 135 mW but a 4 ohm speaker amp would have to be rated at 9*9/4= 20 Watts to power them.

2. Hifiman HE-500 - my test show they need about 7Vrms but these are at 38 ohms = 1.3 Watts. In speaker terms thats about 13 watts voltage swing.

3. Grados - these are tuff, because the are very sensitive/efficient headphones. You really need an amplifier with 60,120,180hz below -80db at 0.4Vrms out. This is a real challenge for us tube guys, but it must be dealt with as there are other headphones sensitive like this. A black black background is a must have not a nice to have.

4. You want your heamp to be flexible, if you meet the above three targets you won't be locked in to a certain model which keeps things fun.

5. We want an SET with a DHT driver and a DHT transformer coupled output, because it just plain sounds good.

Now my current SET headamp is a "space age" tube SPUD SET. I've used 5842, D3A, 6c45p, etc. Simple basic SET with a single tube SSHV, separate chassis, etc. It has a true black background, but these tubes have difficulty reaching high voltage swings, and they also have a harmonic distribution that I hope to improve upon with an all DHT headamp. So the goal is to beat the space age race tubes with an all DHT amp. Of course I have reservations about leaving the simplicity/purity of a single tube but it is required in order to power headphones like #1 & #2.


Today we have newer sand circuits that help with #4. Mainly no iron (choke/transformer) can be allowed within 3 feet of the output transformer. Just trust me on this one, it is difficult if not impossible to have a truly black background with any iron near the output transformers. This means no chokes no IT coupling, even inductors in a DAC can be an issue, thats the sort of sensitivty we are talking about.


The new sand is : SSHV2, Gyrator-mu output, and iko/rods filament. Not only do these great new circuits help with #4 but they help with cost without sacrificing SQ and they keep the amp from turning into a 100 lb behemoth.


Now to the topic at hand as we now have our design goals. And I will start by posting the first all DHT headamp designed by Frank Cooter several years ago, the idea has been around for a while.
 

Attachments

  • frankCooterDHTs.jpg
    frankCooterDHTs.jpg
    272 KB · Views: 4,956

cotdt

Member
2005-10-30 12:57 am
An all-DHT setup will sound amazing, but not easy nor cheap, maybe too expensive on an engineer's salary. 2 stages would work for Grados and Denons but not for some other headphones. You really need 3 stages.

In the schematics, replace 112A with something available like 26 or 4P1L, and 71A with 45. Add a 300B output stage. Replace resistor loading with CCS loads. Will sound brilliant. One day I shall build one.
 

regal

Member
2004-01-04 8:41 pm
MD
Can you include the AKG K-701, 62 ohms, in the debate? I have a pair.

Andy

Those are easy, what I was trying to say is that if you can do #1, #2, & #3 the rest really just fall into place with the right transformer.

Also can you include the Sennsheiser HD600.

Felipe

Same as above, from a SET perspective these are really like #3 maybe with a resistor in parallel depending on how our OPT likes to be loaded.

An all-DHT setup will sound amazing, but not easy nor cheap, maybe too expensive on an engineer's salary. 2 stages would work for Grados and Denons but not for some other headphones. You really need 3 stages.

In the schematics, replace 112A with something available like 26 or 4P1L, and 71A with 45. Add a 300B output stage. Replace resistor loading with CCS loads. Will sound brilliant. One day I shall build one.

The schematic I posted is a starting point for the discussion, circa 2006. We diyers have come a long way since. Its a discussion opener thats all.

I do disagree that three stages are necessary. a single 4p1l can drive a 2A3 into clipping with a modern DAC as source. Sims are showing that #1 & #2 can be driven with possibly even a single 4P1L. Gain is an issue but any gain higher than the bias of the output triode is wasted gain, especially with a mu-follower gyrator load on the driver.

Going from the "purity" of a spud, I would really not want a 3 stage tube amp, it may help a little with #1 & #2 but it would hurt #3.
 

cotdt

Member
2005-10-30 12:57 am
You are right, we should use 2 stages of 26 or 4P1L, followed by 2A3 or 300B. We want the gain without resorting to pentode mode to get it.

More stages would only increase transparency by cancelling distortion. Trust me on this one, I've tried EL84 driving EL84 and it measured better and sounded better than being driven by anything else I've tried. The only requirement is that a good coupling capacitor is used. The tubes themselves are very linear.
 

regal

Member
2004-01-04 8:41 pm
MD
I agree that distortion cancellation is a bit unlooked with SET's. I've simmed driving the 2A3 with a lot of tubes and this is getting off topic but the 6n6p beats anything by about 20 db, yet has more distortion on the 2A3's grid than a D3A or 6c45pi. 2H cancellation is the only explaination.

But I guess I want to keep this amp simple and relatively cheap. I am completly open to ideas but think it best kept a 2 stage amplifier. And I think with the right OPT a 2A3 is too much.

A 4P1l is really what I am leaning toward as an output tube. It can handle the 1W output which covers nearly all orthos except the HE-6 which is probably not worth designing for (its a specialist amp.)

For the driver the choice would be another 4P1l for gyrator-cap coupling, or alternatively a #26 with its lower plate voltage would allow a DC coupling to the 4P1l.

Now honestly I am still "open" to the idea of a 2A3 but I think a 300B is too much, powering it and the size of the OPT would degrade performance for the sensitive/efficient phones like Grados, Senns, modern AKGs.

The real secret to the whole thing is the right output transformer. Headphones like Grados/Senns/AKG's will lose low level signals with a big 4 lb transformer. We are talking about the first milliwatt with these types of headphones. I hope this thread brings a few output transformer experts out of the woodwork ;)
 

cotdt

Member
2005-10-30 12:57 am
I thought your original goal was to include the ability to drive orthos? Otherwise, I think 4P1L driving 4P1L should work great. It is not clear what is the best operating point for this tube, however, so it's best to pick a topology that's easy to make bias changes with.

An 8k:16 transformer would make sense, but it might be easier to get a 5k:8 transformer. Headphones don't have much bass so it's not a big deal. I assume that you will be using a subwoofer?
 

regal

Member
2004-01-04 8:41 pm
MD
I thought your original goal was to include the ability to drive orthos? Otherwise, I think 4P1L driving 4P1L should work great. It is not clear what is the best operating point for this tube, however, so it's best to pick a topology that's easy to make bias changes with.

An 8k:16 transformer would make sense, but it might be easier to get a 5k:8 transformer. Headphones don't have much bass so it's not a big deal. I assume that you will be using a subwoofer?

No subwoofer, this is a headphone amp, orthos have good bass response if thats what you like, in fact my Hifiman an 5LE's probably have better bass definition down to 25 hz than any subwoofer I have ever heard, its quite remarkable really. A 4P1L 5k:38 should have enough power to drive orthos. A good operating point for the 4P1L is roughly 220V 30 mA. Remember if we exclude the HE-6, all we need is 1W for the best modern orthos.

I think the trick to the whole thing is a custom wound output transformer. One that takes advantage of the lower winding ratio which always equals better fidelity, better power transfer, but without detail killing secondary DCR.

We do have to consider output impedance, but not nearly as much as a speaker amplifier.

At some point in the 70's all the speaker and amplifier makers standardized on the 8 ohm impedance. Shortly after it was convenient to add a 120 ohm resistor in series to give a headphone tap. So for a long time headphone manufactuer's "tuned" their drivers response to a 120 ohm tap.

Then Grado decided to change all that and found a hybrid preamp called the SHA and tuned their drivers to around a 10 ohm output impedance. Then AKG/Senn followed with opamp output impedance levels.

Orthos even in the 70's were "immune" to all that. I believe they seem to sound the same regardless of output impedance I imagine because of their completly flat impedance curve.

Ideally what I would like to see come out of this thread is a design with readily available DHT's, SSSHV power supply, Iko/Coleman regs, gyrator, so as to leave enough in the budget for a group buy for a custom output transformer that takes advantage of headphones rather than work against them.

I know enough about output transformers to know that their is potential to do better than an 8 ohm secondary, hopefully if we can all settle on at least the output tube we could catch a winder's interest to really design a nice OPT for headphones.
 

tomchr

Member
Paid Member
2009-02-11 12:58 am
Calgary
www.neurochrome.com
How about using line transformers? They may have issues with the DC plate current, but aside from that I'd think they'd be about perfect.

A 10k:600ohm transformer would present a 5 kOhm load when loaded by a 300 ohm headphone.

If you are targeting a range of headphones with varying impedances, I think you'd be better off with a cathode follower output. Unless you have multiple taps on the OPT, the plate load on the output tube will vary all over the place depending on which phones you plug in.

I think the main challenge with a headphone amp will be to get it quiet. Dead quiet. Be careful with the layout and use regulated supplies (both filament and B+). You probably don't want much if any gain as this amplifies the noise. You do not need power. Most headphones are very efficient. Take the Sennheiser HD600, for example. 97 dB SPL at 1 mW. I usually listen at about 85 dB SPL for critical listening, 100 dB when I really crank it. So I wouldn't need more than 2 mW with typical listening level being around 50 uW (yes, micro-watt).

So let's just say you design for 10 mW to have some margin. Into 300 ohm, that's 1.73 V RMS. 2.45 Vp. That's a tad higher than 0 dBFS on many CD players, so basically, you need a low-noise line stage that can source enough current to drive a low-impedance headphone.

~Tom
 
I'm thinking 4P1L > 4P1L. The operating points are quite well established by now - see the 4P1L SE thread. They can both be used in filament bias - no cathode bypass caps. I'm also thinking transformer coupled stages - I use a Hammond 126C which works fine. Alternatively some kind of direct coupling, but hopefully without a cathode bypass/ultrapath cap.
 

regal

Member
2004-01-04 8:41 pm
MD
How about using line transformers? They may have issues with the DC plate current, but aside from that I'd think they'd be about perfect.

~Tom

The problem is too much DCR, it affects the frequency response of headphones like Grados, Denond, AT's, modern AKG's, etc. Most line output transformers have a DCR over 100 ohms.

I think we should focus on the tubes first then hopefully Bud P or one of the other few remaining OPT experts can point us in the best direction.

For the output tube I think my vote is the 4P1LE, its cheap readily available, very linear, and capable of over 1W output. It also has a lot of gain which makes it easier to reach the overall gain we need with a driver like the #26 or another 4p1L.

The driver is a bit of a tough choice. Going off Cooters design he wa very adamit about DC coupling. If we use a #26 the plate voltage is low enough to DC couple and still use an SSHV2. The with DC coupling is the large bypass cap on the output.

Now the biggest question I have is something I have pondered since I first so that schematic is how to get rid of the bypass cap and DC couple. Negative supply for the input tube?

Or with a modern Gyrator Mufollower on the driver maybe we don't need to worry about cap coupling to the output stage? Then we could filament bias both the tubes.

One thing I learned when I built the SSHV SET 5842 spud headamp was that a big part of its magic was no big caps in the signal path. I have an LED bias and the SSHV, no input cap and no big electrolytic on the cathode, it makes a very clear absolute honest sound.
 

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
Most line output transformers have a DCR over 100 ohms.
And most speaker level OPTs have secondary DCR of a few ohms.
FWIW, I was almost able to get enough volume to drive my Sennheiser cans out of the triode strapped 4P1L SE into a 32:1 OPT. An 8R across the output in parallel with the cans. Seems to me with a nominal 2.5K:8 you'd be at about twice the output voltage already. How much gain would you need to get enough voltage for headphones?

As was said earlier, you mostly want a line amp with lowish output impedance.
 

regal

Member
2004-01-04 8:41 pm
MD
I know enough about transformers to know it largely emperical (meaing there are a few gurus with real tribal knowledge.)

So I agree with line level sized transformer just not the 600 ohm secondary, 5k:40 would put us right where we need with over all system gain of ~80 even lower gain with a solid 2Vrsm source.

The other option we have would be two 10 ohm secondaries where the output could be "balanced" for the orthos/senns, this would be 5k:40 balanced, and then 5k:10 for singe ended to use with the sensitive phones.

The fidelity of the transformer is better with smaller cores, I think the standard 5W core is too big. And the fidelity of the transformer is better with lower turns ratio because the coupling is better.

What we want is technically feasible its just finding someoe with the background to pull it off.
 
And most speaker level OPTs have secondary DCR of a few ohms.
FWIW, I was almost able to get enough volume to drive my Sennheiser cans out of the triode strapped 4P1L SE into a 32:1 OPT. An 8R across the output in parallel with the cans. Seems to me with a nominal 2.5K:8 you'd be at about twice the output voltage already. How much gain would you need to get enough voltage for headphones?

As was said earlier, you mostly want a line amp with lowish output impedance.

You don't need to load by 8 Ohms. Just solder there 36 Ohm resistor, or even 62 Ohm, it will be enough to protect transformer from blowing by flyback voltage, while distortions much lower than when loaded on 8 Ohm.
 

regal

Member
2004-01-04 8:41 pm
MD
You don't need to load by 8 Ohms. Just solder there 36 Ohm resistor, or even 62 Ohm, it will be enough to protect transformer from blowing by flyback voltage, while distortions much lower than when loaded on 8 Ohm.

Under stood, but we want to take advantage of a lower winding ratio. The lower the winding ratio the higher the fidelaty, this is because of the way the secondary is wound with the primary.

It also gives better power transfer, we can deliver much more power per gived plate swing with a 5k:32 ohm transformer than a 5k:8 transformer. This off course means less distortion.


I am also kicking around parafeed, then we could use a nickel permalloy OPT, what are the thought of parafeed vs single feed for this project ?
 
One thing I learned when I built the SSHV SET 5842 spud headamp was that a big part of its magic was no big caps in the signal path. I have an LED bias and the SSHV, no input cap and no big electrolytic on the cathode, it makes a very clear absolute honest sound.

I experienced the same so completely agree with you: LED bias and the SSHV, no input cap and no big electrolytic on the cathode.
 

regal

Member
2004-01-04 8:41 pm
MD
Yes that is what I use on my spud, 2xmitsu 5AR4's feeding a CLC choke to the 2 SSHV's.

SS regulation is a must imho for a black background with phones that have specs that read like Efficiency @ 1mW (dB SPL) 32 ohms (Beyer T5's.)

Its just too costly to get ripple so low for phones like this without SS regulation, and the SSHV is very transparent/natural sounding. Comparing it to a maidia for expample.

I got out my quad of 4P1L's, boy they are odd looking tubes. Trying to find a western equivalent, spec wise I think they are sort of a cousin of the 306A/307A ?