Measuring transformers

[SimonC]

The 30k may well be from the pc that you are using to measure, or any other switching supply in the region.

OK, thanks Nick. I might try running the laptop off the battery tomorrow. Would be good to get to the bottom of it. Might become a problem in the future when I'm measuring phono stages
...

I've just had a similar problem, the culprit was my lads fancy multi-colour led lit keyboard...

[Mike H]

This is the frequency response of the 6i6 in a loop-back test. I guess they give it a soft roll-off to sound better?
I think this means the SNR will be worse outside the audio band?
I wonder if the roll-off could be modified....

Looks pretty much like every other transformer I've checked the freq response of (including LTspice models) - the LF roll-off is due to the core being too small at that freq, the HF roll-off is simply because of the rising reactance of the windings, adding series impedance. Also there is inter-winding capacitance, so you can get a low-pass LC filter effect as well. But I see nothing unusual about that plot except that the HF is quite high.

[Max N]

Looks pretty much like every other transformer I've checked the freq response of (including LTspice models) - the LF roll-off is due to the core being too small at that freq, the HF roll-off is simply because of the rising reactance of the windings, adding series impedance. Also there is inter-winding capacitance, so you can get a low-pass LC filter effect as well. But I see nothing unusual about that plot except that the HF is quite high.

Hi Mike
I think the plot you're looking at is the roll-off of my sound card. It was a loop-back test, the transformer wasn't connected.
This was the plot of the transformer - flat from 4Hz to 98kHz. I need to figure out how to extend the measurement below 4Hz to see where the LF limit is.....

[Nick]

You also should drive the transformer into its expected load and from its expected source impedance to see how it will behave in actual use.

[Max N]

You also should drive the transformer into its expected load and from its expected source impedance to see how it will behave in actual use.

Yep, that’s next on the list. (As I said, I hope to measure everything that’s on that Jensen data sheet)

The source impedance of the soundcard is quoted as 94 ohms balanced, and I’ve done a quick test to verify that. So currently my setup is a long way from how I have been using these transformers (mainly driving them with 5687 and similar). One of the motivations for this measurement exercise is to understand how to get better performance from the transformer.
My thoughts are different valves, or parallel valves, or some kind of buffer or follower....
But first I need to establish exactly how low the impedance needs to be - or more accurately, where is a good compromise.....

[simon]

Nice thread Max, looking forward to reading more

[Max N]

Nice thread Max, looking forward to reading more

Cheers Simon, it’s taking me a bit longer than I’d like. Hope to have some more measurements soon.

[Max N]

As Nick suggested, here is some data showing the effect of source impedance. I haven't investigated the effect of load impedance yet, but the 6i6 is quoted as 52kohm which I guess is not too unrepresentative of the grids of the following stage as long as we stay away from A2. I will do some tests later at much lower impedance to simulate driving grid current.

I wanted to extend the test to lower frequencies but I found that I had to drop the sampling rate of the 6i6 down to 48kHz, which then meant an upper frequency of 24kHz. So what I have ended up with is two separate sets of curves - first here is the low frequency results from 1Hz to 20kHz.

The upper family of curves are the amplitude response, using the Y-axis graduated from +5dB to -20dB.
The lower curves are the phase response, using the other Y-axis graduated in degrees.

Red is 0 source impedance (I think the measurement method eliminates the impedance of the 6i6 because of the way the reference signal is derived)
Green is 680 + 680 ohms
Blue is 1200 + 1200 ohms
Magenta is 1800 + 1800 ohms (roughly representative of driving the IT with one half of a 5687 per phase). This looks to be about 1.5dB down at 3Hz

[Max N]

And this is the higher frequency result, from 70Hz up to 90kHz.
Same colours/impedance as before.
Not especially informative because of the upper frequency limit of the soundcard. The 1k8 + 1k8 (magenta curve) is starting to dip at 90kHz, about -0.5dB compared to the response at 3kHz. I will have to use a signal generator and scope to get a -3dB point for each source impedance.

Next step I will drop the load impedance to simulate A2 operation of the following stage......

[Max N]

Here is the effect of load impedance. Source impedance is the same for both red & blue measurements - about 1K8 plus 1K8 or 3K6 balanced.
The red line is loaded by the audio interface (52K), which gives about a 1dB drop.
The blue line is loaded with 3K6. The source impedance and load impedance act as a potential divider, so we lose half our signal, hence we are a further 6dB down (roughly)
Actually the bandwidth of the transformer itself is still pretty good, within 1dB from 4Hz to 90kHz.
So if we wanted to drive a purely resistive load of 3K6, we would be in pretty good shape. But what if we wanted to drive a push-pull pair of output tube grids into A2....?

I've chosen these impedances to be somewhat representative of that situation - say a pair of 5687 driving a pair of 211s into A2. The problem then of course is that for most of the voltage swing, we would have a load impedance of maybe 100K (depending on the grid circuit resistance of the output stage). As we swing into positive grid volts, the grid starts to conduct. I seem to remember that when I looked at 211 datasheets for grid current vs volts, the grid drew about 25mA at +50V, equivalent to about 2KOhms.

So we have a load which transitions from 100K to 2K (I'm sure I could rig up a dummy load to simulate this with some resistors and a diode but I need to think about exactly how to do it). We end up with a 1dB loss for most of the signal swing, transitioning into a 6 or 7dB loss. The end result will be distortion - I can visualise what it will look like, the peaks on one side of the swing will be flattened. The more we drive into A2, the more flattening.

Anyway, this is all a bit of a digression, because actually this is just the effect of grid current and not really a characteristic of the transformer. The transformer itself seems happy enough driving a 2K resistive load.

However, if I want to use this IT to drive a pair of output valves into A2, I will need a low source impedance. Any suggestions on how to do that?
I could parallel a pair of 5687, but that will only get me down to 1K.
I have some E810F and some E55L which I could run as triodes to get me lower than the 5687.

Or I could use some kind of follower or buffer......

[Max N]

In the meantime, back to measuring the ITs. The next thing I want to investigate is the effect of DC imbalance...I haven't figured out how I'm going to do that yet.......

[Max N]

So, more than a year later....

If I ground either the primary centre tap or the secondary CT, the 30kHz noise is gone. Looking at these traces, I think we can say that this transformer is not picking up any measurable noise, down to about 130dB below a 1v signal. Since these ITs are intended to drive output valve grids, they will be carrying at least several volts of signal. This seems like quite an impressive result for unshielded transformers?
I guess it's because the windings are bifilar and/or very well balanced, so any noise signal induced in the windings will be common-mode and not seen by the balanced input of the soundcard. If I use well-balanced pairs of valves then I should see the same common-mode rejection in-circuit?

I've been thinking about this . I agree with myself about noise signal induced in the windings. But since then I have run into problems with noise induced in the core - and that doesn't get rejected. These transformers are quite good at picking up stray magnetic flux. I was using EI transformers to drive some heaters - had to switch to toroidal. And then there were intermittent HF bursts - I noticed it coincided with the lights on the power line adapters. Switch mode power supplies were another culprit.

[Max N]

I don't think I ever did get around to measuring the effect of dc current.
But I did do some measurements on the effect of source impedance on LF distortion

The 0 + 0 line is the transformer being driven by the 6i6 sound interface (94 ohms). And the 680 + 680 is effectively 774 + 774 etc.

I think I knew that driving a transformer from a low impedance improved its performance, but going through this exercise has driven it home.

[Max N]

I think that this thread has more or less run it's course now. If I do make any more measurements that are purely indicative of the transformer performance I'll post them here.
Armed with this better appreciation of how to get the most out of these transformers, I decided to have another go at a PP -> IT -> PP amp.
I'll document what I've been up to in another thread.