Not necessarily all headphones, since there are high impedance headphones as well. The [theoretical] loss is much lower when you are driving HD800's with an impedance of about 350 Ohm via Lundahl LL1689 OPT with a secondary DCR of about 15 Ohm (depending how they are connected...)
But I fully agree that choosing really good output transformers is the way to go for either route...
Yes you are right. The only problem is they are relatively heavy and bulky because of the power handling at low frequency.
I have already played with the LL1627 and LL1623. For example, if you use the LL1627 as 1.2K/8R transformer with 32R headphones you will get about 4.2K primary impedance. Having 18H low end cut-off (-3 db) will be around 6-8 Hz with valves like the 2A3 and the 4P1L. Insertion loss will be very low around 0.2 db or even less with 4P1L.
The same LL1627 can be used as 0.65K/8R for 64R headphones with similar results.
I am thinking about commutation of secondaries (quality paid with complexity) for headphones with an order of magnitude difference in impendance (32 and 300Ω, plus 8Ω for small speakers) vs. a resistive ballast (simplicity paid with distortion). So just for a test, I put together an xls for LL1627, LL1623, LL1620, LL9202 that computes impendances of all possible combinations of their coils and the headphones of interest (32, 38, 52, 64 and 300Ω). For the tube I want I need 2500~3000Ω.
LL1620 (without 52Ω), LL1623 (without 300Ω) and LL9202 (without 300Ω) can have secondaries easily commutated to accept all interesting headphones and 8Ω speakers. And 300Ω can be commutated with a shunt resistor to match it to one of the other values (i.e. 64Ω).
So I do not see a desparate situation here, if small extra complexity (and a microcontroller to run all those relays and R-sensors) is not a taboo.
Can you post the spreadsheet?
One thing I am not sure of is if we think about a 3K:38 transformer with a single secondary, think about how much lower a winding ratio this is than a 3k:8 transformer. As we all know the lower the winding ration the better th quality. With the LL's they use multiple secondaries so I think of that benefit or potential opportunity is lost?
In principle is not a good deal because it is already difficult to keep distortion low even with a proper OPT. We are dealing with a situation where you need very little power for high SPL (i.e. headphones) in comparison to loudspeakers.
Anyway there is nothing wrong if you try. At least you can be sure!
With Lundahls you can have several optimal combinations because the single section terminations are brought out. You just need to do this for the secondary. The primary sections are meant to be all in series for minimum capacitive load.
The lower turn ratio is not better a priori. It depends on the application but generally speaking is typically worse already when dealing with impedances in the Kohm range. If you compare a 5K IT with 1:1 ratio and a 5K/8R OPT with identical geometry except the serial-parallel connections of the secondary the latter is a lot better because the leakage inductance and the stray capacitances of the secondary are neglectible.
Attached. But there may be silly mistakes in there, so verify before making any decisions.
I am looking at a rather narrowly specified design: single-tube EML AD1 (grid driven directly by high-voltage I/V DAC, at -45V) at 250V on anode and 60mA, as it is meant to be, which gives about 4W at 8Ω load. Having read your initial requirements (in general a good match with what I want) and the following discussion, the transformer for a widely varying Z loads is the main complication, and it looks like multiple secondaries with some combination of secondaries' commutation and gentle shunting (for odd loads) may accomodate all potentially interesting headphones plus 8Ω speakers at a cost of relatively small increase in overall complexity (added Reed relays and a microcontroller).
Hello
This may be something to add to the conversation.
A couple or 3 years ago I had Edcor design a transformer for 300 ohm HD600 headphones. The requested transformers were specified as 3 watt, 50 ma idle current, and 5K input : 300 ohm output with a 40% tap (to experiment with ultra-linear connection). I ordered 10 to avoid the design fee. The Edcor model number is GXSE3-300-5K. If you want to try a pair for your prototype; they cost me about $20 each back then.
DT
This may be something to add to the conversation.
A couple or 3 years ago I had Edcor design a transformer for 300 ohm HD600 headphones. The requested transformers were specified as 3 watt, 50 ma idle current, and 5K input : 300 ohm output with a 40% tap (to experiment with ultra-linear connection). I ordered 10 to avoid the design fee. The Edcor model number is GXSE3-300-5K. If you want to try a pair for your prototype; they cost me about $20 each back then.
DT
Pieter's take on a DHT headphone amplifier.
http://www.diyaudio.com/forums/tube...-all-dht-headphone-amplifier.html#post3368395
http://www.diyaudio.com/forums/tube...-all-dht-headphone-amplifier.html#post3368395
On the lundahl datasheet they show only one primary connection "choice" for SE. They have the first two primaries in series connected to the B+ and tube then the other two primaries connected to each other. I'm not sure what to call that is what they are calling out all four primaries in series, seems as if they don't recommend anything else.
This seems like a good place to ask this question even though it's not DHT: I've been thinking about trying to use a 6S31B tube in a headphone amplifier application (30-70 ohm range). I was thinking a regular old 2 stage GC to CF with an output transformer as the CF load. There are a couple OPT's that might work for this since they were designed for a replacement in a solid state CF-output application (Lundahl and Cinemag CM-96731). Does anyone think this idea has any promise? I know I could use a 2200uF output cap instead of an OPT but...
It depends on the actual model. For istance the LL1623 has a primary made of two halves and each half has two sections (so 4 identical sections in total). They really are two halves because there are two separate coils, one on each straight of the 2C core.
Other models, like the LL1664, have just two halves, one for each coil, with no further sectioning. There is a intermediate tap for UL connection on each half. All the variants for different loads are given by the secondary sections arranged in all the possible combinations.
OPT primaries, where accessible, are usually meant to be in series otherwise forget a good frequency response. Stray capacitances would be to high for parallel connection. More likely the tube could see a difficult load to drive and the merit factor Q at the OPT self-resonance could be non-ideal resulting in a nice sharp peak which you can see on the oscilloscope as persistent ringing superposed on square-waves. If the Q is high, say more than 1, the ringing will be marginally damped or not damped at all. Maybe for zero feedback is not a big problem if the resonace is well above 20KHz, with Fb it wouldn't be desirable at all. In principle tubes don't like high capacitive loads.
Last edited:
Does that apply to a parallel connection of secondaries? There are 8 of them in LL1623, and to accomodate a range of Z's I need 3, 6, 7, or 8. Loose secondaries are not recommended by the application note, parallel connection is. Will there be any negative effect in sound quality due to that? (excluding all possible issues with commutation for simplicity) Looking at commutation circuit, I think of parallelling all the unused secondaries. For example, for 38Ω six secondaries must be in series, which leaves three that must be paralleled. Any comment on that?
Also, if an odd Z is shunted gently to match it to the nearest value, i.e. 300Ω to 64Ω, with the same power in the secondary, would there be any negative effect in sound? For example, 4W maximum output at 64Ω (16V) splits as 0.85W at 300Ω headphone and 3.15W at 81Ω shunt -- more than enough power for Sennheiser HD6x0, or a total loss of hearing, so the useful maximum power level would be lower by a multiple, at least by 2.
Yes this applies to the secondary too but you have to divide by a factor equal to the turn ratio. It is a much smaller effect and in practice is negligible. More importantly you have to consider that the transformer has been designed to work like that (i.e. primaries always in series and secondaries in series-parallel combinations).
That depends on the linearity. If you can get 2-4W with very low distortion, say similar to what you get when your Pout is 0.4-0.8W without shunt, then is fine. You can try. You can see in the other discussion on 4P1L that you can get up to 2W with only 0.6-0.7% from a PSE with 5K OPT (basically 10K per valve). This would be fine, I guess....