Curious issue with replacement coupling transformer

[yestertech]

I've been awaiting the arrival of some replacement transformers for a stereo single ended 300B amp.
The driver stage (5687) is coupled to the output valves via a 1:1 transformer, wired in its anode circuit. One of these had developed a short primary to secondary which produced incessant crackling as HT was being leaked into the grid circuit of the 300B.
After some research I tracked down a contact within the original amplifier maker and from this came the original specification for the coupling transformer.
I subsequently managed to contact the transformer manufacturer and using the stated spec. 2 new transformers were wound. ( the makers recommended replacing both to preserve the channel matching etc. )

Yesterday I was able to fit these exactly as the originals. Wiring checked, I powered up the amp and all seemed fine.
Any sense of success vanished immediately on a listening test. The sound was very dull with most of the HF response above about 5Khz missing. Both channels the same.

A quick frequency sweep from input to LS out revealed a most peculiar response, with HF roll off starting from around 4 Khz and a massive LF lift by 100Hz.

I can't fathom how this can be, given the replacements were produced from an identical specification.
With power off, I ran the test again just from transformer primary to secondary (albeit in circuit ) and its broadly the same. HF loss and massive LF peak.

As a sanity check, I repeated the primary:secondary test on the removed working transformer and this gave a pretty flat response as expected.

Could reversal of the winding 'polarity' on primary or secondary somehow cause this ?

I'm baffled

Andy

[Ted Kendall]

I think the obvious inference is that the original spec has not been followed, either with regard to materials or construction - the lay-up of the windings is critical to balance inter-winding capacitance with winding inductance. Get this wrong and attenuation or resonance may result. Core material also needs to be as specified, of course.

[yestertech]

Thanks Ted - In my mind that would seem to be the logical conclusion.
However, it would be VERY disappointing given they were supposed to be working from the amplifier manufacturer's numbered specification.

A.

[Ted Kendall]

Quite so. I suppose you could put the matter beyond doubt by doing some measurements of inductance and capacitance and comparing those with the working original or, in the last resort, dissecting the faulty original to determine construction details. Perhaps Ed Dinning could offer some advice?

[barrymagrec]

Are you absolutely sure that the same lamination material as specified was used?

I repeated the primary:secondary test on the removed working transformer and this gave a pretty flat response as expected.

What were the drive and load impedances in that test. In circuit I guess they will be very different.

[tomv893]

Originally Posted by yestertech

I've been awaiting the arrival of some replacement transformers for a stereo single ended 300B amp.
The driver stage (5687) is coupled to the output valves via a 1:1 transformer, wired in its anode circuit. One of these had developed a short primary to secondary which produced incessant crackling as HT was being leaked into the grid circuit of the 300B.
After some research I tracked down a contact within the original amplifier maker and from this came the original specification for the coupling transformer.
I subsequently managed to contact the transformer manufacturer and using the stated spec. 2 new transformers were wound. ( the makers recommended replacing both to preserve the channel matching etc. )

Yesterday I was able to fit these exactly as the originals. Wiring checked, I powered up the amp and all seemed fine.
Any sense of success vanished immediately on a listening test. The sound was very dull with most of the HF response above about 5Khz missing. Both channels the same.

A quick frequency sweep from input to LS out revealed a most peculiar response, with HF roll off starting from around 4 Khz and a massive LF lift by 100Hz.

I can't fathom how this can be, given the replacements were produced from an identical specification.
With power off, I ran the test again just from transformer primary to secondary (albeit in circuit ) and its broadly the same. HF loss and massive LF peak.

As a sanity check, I repeated the primary:secondary test on the removed working transformer and this gave a pretty flat response as expected.

Could reversal of the winding 'polarity' on primary or secondary somehow cause this ?

I'm baffled

Andy

Can you unwind the faulty transformer to establish its winding profile, if you can unwind it, ask the manufacturer about their winding profile and see if it matches.

[kalee20]

The choice of core lamination material won't affect the HF end of the response curve (it has impact at the LF end).

I'd say the issue is in the winding configuration - too much self-capacitance!

[Leon Crampin]

I would disagree. If an audio transformer is wound with "mains transformer" laminations, the iron loss at high frequencies will be considerable - it varies with the square of frequency as I remember.

Before you strip the transformer, measure the lam thickness.

Leon.

[yestertech]

Originally Posted by merlinmaxwell

What were the drive and load impedances in that test. In circuit I guess they will be very different.

I tested this with primary fed from a sig gen. output of 150 ohms and secondary unterminated.
The same set up was used to test the new transformer, the only difference being that it was still wired into circuit. ( unpowered )
Primary between HT + and valve anode, secondary between bias pot output and 300B grid.
(this is exactly as originally wired )

A.

You could have an AC conductive circuit twixt the bias pot and HT+, try reversing one side. The anode and grid will be essentially floating (more so with the valves out), try bias and HT+ as the ground conexions.

[Radio Wrangler]

Because of the use of 300Bs, single ended topology and transformer inter-stage coupling, I'm guessing that this amplifier doesn't use any negative feedback, so we aren't too worried about polarity reversal?

David

[kalee20]

Originally Posted by kalee20

The choice of core lamination material won't affect the HF end of the response curve (it has impact at the LF end).

Originally Posted by Leon Crampin

I would disagree. If an audio transformer is wound with "mains transformer" laminations, the iron loss at high frequencies will be considerable - it varies with the square of frequency as I remember.

Ah, but you're thinking of eddy current loss variation with frequency alone. Don't forget that the flux density falls with frequency, for a given transformer operating at the same maximum voltage. At 10kHz, the AC flux density will be only 1% of the AC flux density at 100Hz, and this 'wins'.

Below is loss characteristics of 0.3mm silicon-iron (source: Telmag 'C' core data book). You can see that at 30Hz and 100mT, loss is 7mW/kg. At 10x the frequency (300Hz) the losses are 140mW/kg, not 700mW/kg. It's a 14x rise.

By contrast, at 100mT 300Hz the losses are 140mW/kg (as already stated); at 300Hz 1T we're looking at 6W/kg which is 43x as much.

So (although these curves don't extend to 10kHz) you can see that the rise in losses for a frequency increase is more than offset by the fall in losses due to flux density drop.

[yestertech]

Originally Posted by Radio Wrangler

Because of the use of 300Bs, single ended topology and transformer inter-stage coupling, I'm guessing that this amplifier doesn't use any negative feedback, so we aren't too worried about polarity reversal?

David

That's correct - there is no feedback used.
A.

[Ed_Dinning]

Hi, some good answers above.

Well worth trying to open up the transformer if possible without damage, and see how many sections it is wound with and if it is of a similar construction.
This type of transformer is very sensitive to interwinding capacitance, much more so than the simple ones used in 1920's valve sets.
Note that some of the early Ferranti I/V transformers incorporated small capacitors to tweek the response of the unit, I can't remember if these were in series, parallel or across the windings, but they are clearly shown in "The True Road to Radio".

The suggestion of reversing the phase of one winding is worth trying as there is no feedback applied. If this makes the part of the (say) primary earthy where it is next to the (say) sec, then that could well reduce self capacitance

Ed

[Radio Wrangler]

I'm afraid, Andy, it looks like you're locked into retracing all the reasons R-C coupling replaced transformer coupling at audio. The only reason there is transformer coupling in that unit is a certain sort of boutique-y feel good factor.

As Ed says, with a single-ended design, interwinding C coupling may be somewhat different with one winding or the other reversed. Nowadays we expect amplifiers to cover a few more octaves than old radio set interstage transformers would let through. One key to doing broadband transformers is to run them in low impedance regimes, but interstage transformers are in high impedance regimes. Because these people think low component count = simplicity, and simplicity is next to godliness, then they are squeezing every at bit of gain they can get, and that means higher stage load impedances.

It's frustrating having to fight such self-imposed handicaps, when there are good cures possible for use in amplifiers based on sound engineering.

You might have to tell him that it really needs a part from the original manufacturer. Only that way would he get matching responses in the two channels.

David

[kalee20]

Following on from David's post above, and re-reading the kick-off post, the transformer is 1:1 (so it's not providing any voltage gain) and it's feeding a single-ended stage so it's not providing push-pull drive either... what's to stop you just using RC coupling anyway?

Or, if plenty of HT on the driver stage is needed to secure a good level of drive before clipping, choke loading? With just one winding rather than two, self-capacitance will be reduced, all other things being equal.

Low DC resistance in the grid circuit almost certainly won't be a 'must-have.'

[Edward Huggins]

To avoid any more time consuming struggling, the tough conclusion here is that this will need an original replacement part from the manufacturer - nor to go to RC coupling. That use of an intervalve transformer smacks of exotica and a desire for product differentiation.

choke loading

You could put loads of different ones in series, shorting out some for tone control.

[Ted Kendall]

Originally Posted by yestertech

the makers recommended replacing both to preserve the channel matching etc.

Illogical, captain - it seems they are saying they can't guarantee the spec of their own transformers, never mind anybody else's. I don't think you have much to lose by stripping the faulty transformer and rewinding it in the same fashion. Are the manufacturer's replacements unobtainable or merely prohibitively expensive?