Aikido as a headphone amplifier (Output transformer groupbuy)

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Administrator
Joined 2004
Paid Member
Hi Bas,
Just wanted to let you know that the transformer arrived a few days ago at my doorstep.. I have not yet started the evaluation to see where we stand, but I am very concerned about leakage inductance, based on the way it is constructed.. Also I would like to know what the number of turns on the primary are and the target inductance. Maximum allowable flux density and turns per volt would also be useful information. (This last is not usually specified with audio transformers but I am curious.)

I need to verify what the actual winding ratios of the 3 separate secondaries are in order to determine what output load impedance combinations are possible. If the three windings are all 200 ohm windings this limits us to 200, 800 and 1800 ohms. Should they be different they cannot be paralleled which still leaves us with just 3 possible output impedances. I will verify..

Looking at the spec it appears that they are 200 ohm windings and this gives us only 3 valid combinations 12:1 (200 ohms) all in parallel, 4:1 (1800 ohms) - all in series, and two in parallel with one in series 6:1 (800 ohms) if I got the math right.. No other possible combinations exist.

Remember in an impedance transforming device like an output transformer equal windings in parallel do not change the transformation ratio therefore the optimum load impedance versus the reflected source impedance does not change.

Windings in series add as the square of the winding ratios so two equal windings in series yield an increase of 4X and 3 equal windings in series yields a 9X increase.

Note that you cannot put dissimilar windings in parallel as this is technically the equivalent of partially shorting some of the turns.

edit: major LOL
 
Administrator
Joined 2004
Paid Member
Results of Transformer Eval.

Hi All,
Results of the transformer evaluation look promising, it does look like an R core output transformer will provide adequate performance. Significant design changes are required, however the basic concept looks workable.

I measured a primary L of 11.2H this is adequate for this application as long as the driving impedance is not greater than 600 ohms. (Probably not a problem with the Aikido topology.) No DC!

Secondary L of 100mH, and the inductance of the secondaries was well matched indicating good winding symmetry.

Leakage inductance with all 3 windings shorted was ok at 48.3mH, I would like to see lower.

All was not well however when all windings are not used - this doubled leakage inductance or more.

Bandwidth into a 200 ohm resistive load and driven by a 600 ohm source was very good.

I measured -1 dB point at 79kHz at 3Vrms drive to the primary, and -1 dB point was below the 8Hz limit of my Amber analyzer.

Dcr of the secondaries was 4.4 ohms. and of the primary 264 ohms.

Note that we should loose the electrostatic shield between the primary and secondary in the next round - there is 300pF capacitance between the shield and primary right to ground. Between one of the secondaries and the primary capacitance is about 150pF. Total capacitance theoretically is around 750pF or more.

There are 3 valid ratios and I verified these: 12:1 (all in parallel), 6:1 (any two in parallel with the remaining in series) and 4:1 (all in series)

Note that the impedance transformation for 12:1 is 10K to 70 ohms, NOT 200 ohms as reported. Which is close enough to one standard headphone Z to be useful. Note the impedance transforms as the square of the ratio so 12:1 is 12^2 or 144 times so if the target primary Z is 10K divide this by 144 to get the load impedance that reflects to this primary Z which is about 70 ohms.

The others are ~600 ohms (4:1) and ~322 ohms (6:1)

All of these values are close to some standard headphone impedances, however performance with 8 ohm phones will be limited and I suggest reconfiguration to support 8, 32 and 64 ohm phones and possibly 600 ohms. I can work out the required ratios as needed.

I have not measured distortion yet. I don't expect a problem however.

The existing wiring diagram is not correct and should not be used as a reference. Black and white are the primary connections, not as shown. Most of the secondary wiring configurations are incorrect, only three valid combinations exist.

Please weigh in with your load impedance requirements and if reasonable I will work with Bas and the designer to incorporate in the next round.

I would like a stereo pair for evaluation once we have the revised specs nailed down.
 
Administrator
Joined 2004
Paid Member
Something to note, the leakage inductance will have progressively greater influence on HF response as the load impedance is reduced for any given impedance transformation ratio. LF performance will conversely improve until excessive load current in the primary results in core saturation.

I will post additional results as I do the measurements.
 
Administrator
Joined 2004
Paid Member
Here is some more information illustrating the effects of changing the loading.

Loaded into 68 ohms with all windings in parallel the -1 db point is 33kHz as opposed to 79kHz loaded into 200 ohms showing the effect of the leakage inductance.

Here is what happens with one winding as compared to all three in parallel into 68 ohms, -1dB is now a mediocre 23.7kHz.

I think we need to pay attention the overall level of leakage inductance, eliminating the electrostatic shield will help as will careful interleaving of the primary and secondary windings.

IF we can reduce leakage inductance to 25mH or less in the next round this should be satisfactory.
 
Administrator
Joined 2004
Paid Member
I have been thinking about the rated primary impedance of this transformer since I measured the primary inductance this afternoon. In order for this to be actually true the primary inductance would need to be more than an order of magnitude higher.

We know the inductance is 11.3H and at 20Hz this gives us a reactance of 1.42K ohms. The load the transformer presents to the driving stage is in fact the thevenin equivalent of the parallel reactance which should be very large and the transformed load impedance, to be rated a 10K primary the equivalent impedance should be dominated by the reflected load impedance which implies a reactance of about 10 times greater magnitude in order for this to be basically true, this is the case at frequencies above about 1.7kHz, but not below it.

What this implies is that the source impedance most be a lot lower than the 10K primary rating indicates, I believe it also means that substantial amounts of energy are going to be lost in the winding itself rather than being delivered to the load.

I think in this application a better way to look at this would be to regard it as a step down transformer with certain specific ratios and a specific primary inductance.

I suspect the primary issue is probably due to the relatively low permiability of the rcore.

Another option we have would be to configure this as an auto-transformer, providing specific matching ratios - this would force the user to put the coupling cap on the output side of the amplifier rather than on the hot or cold end of the transformer as previously possible.

The low dcr of the auto-transformer would prevent significant amounts of dc appearing at the headphones in the event of the coupling cap failing.

So we can do a conventional transformer with separate primary and secondaries OR an auto-transformer.

I am currently thinking that 8, 16, 32, 64 ohms ought to be supported and possibly something higher than this as well.

In the conventional transformer these can be supported by two secondary windings with 8 and 16 ohm taps - how they get wired determines the output impedance.

In an autoformer you would simply tap the primary at these points - theoretically getting lower leakage inductance would be easier in this case as there would never be any parts of the winding that were not used. (Although in practice this may not be the case because of the dual bobbin construction used in the rcore transformer.)

Any thoughts?
 
32 (grado) and 300 ohms (sennheiser 580/650) are the 'typical' hp impedances. Many phones intended for portable use go as low as 16 ohms (etymotic er-6i), and on the high end I think that 600 ohms is the limit for 'ordinary' headphones. I believe that 32 and 300 ohms would take care of the vast majority of headphones though.
 
"I have been thinking about the rated primary impedance of this transformer since I measured the primary inductance this afternoon. In order for this to be actually true the primary inductance would need to be more than an order of magnitude higher.

We know the inductance is 11.3H and at 20Hz this gives us a reactance of 1.42K ohms."



How did you measure the primary inductance? I noticed an earlier comment about needing an LC meter with a 200 Henry range I think. One cannot measure the primary L with an ordinary LC meter. They use too small a test signal, all you will see is the low initial permeability of M6 steel.

Have to use a variac and Volt and Amp meters to plot out the primary L curve versus voltage input, which starts out very low and should peak much higher within the operating range somewhere.

All usual OTs behave that way except for special high Mu alloys like Nickel/Iron. Since the normal Aikido does not have very low output impedance like a feedback amp, it will not have a high damping factor to handle the M6 core well, so it might be a good idea to go to a permalloy core material like in interstage xfmrs.

Don
 
Any updates on this?

As a headphone afficionado I can vouch for the wisdom in having 32ohm and 300ohm as the Grados and Senns are amongst the most popular high end headphones out there.

AFAIK most of the high end Sonys and Audio Technicas are 32ohm rated as well.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.