• 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.

OPT resonance Q factor in the HF region - your opinions?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
What would you choose between these two options and why? I'm curious to have your opinions.

1. Optimized Q factor, no apparent ringing, optimized transformer for the schematic and tube, but lower resonant frequency. Of course, roll-off due to Ls or Cp at much higher point (optimal tube and load)
Some numbers:
Ls - 0.69mH
Cp - 2nF
Fres - 135.5kHz
Q - 0,77


2. Slight ringing, higher Q, ugly square wave, but also higher resonant frequency. The parameters in this situation are:
Ls - 1.13mH
Cp - 667pF
Fres - 183.5kHz
Q - 1.7
 
I would have to know the amplifier topology before making a choice.

1. If the amplifier Does have negative feedback, I would probably choose transformer #1. You have to worry about square waves into resistive loads during a Reviewers testing (or a audio club member's testing). You also have to worry about the real world loading from loudspeakers and loudspeaker cables, when using the amp for listening to music.

2. If the amplifier does Not have negative feedback, And is either a single ended triode amp or a push pull triode amp, I would use either transformer. That is because the music sources I use do Not have frequencies higher than 20kHz. There will be No ringing in that application. And sense there is no negative feedback, you do not have to worry about stability. So how about some other criteria for selection of the transformer:
Which transformer is less expensive?
Which transformer has more inductance?
Which transformer has better primary and secondary DCR?
Which transformer has more laminations?

Because I am not making a commercial amplifier, with some Reviewer or audio club member trying to criticize the amp's bench performance, I do not have to worry about some ringing at ultrasonic frequencies.

Some persons will say I need to worry about the phase performance of the amplifier. I have a story about phase: Some early CD players used only one 16 bit DAC, and a switch that alternated the DAC output between the left and right channels. The DAC outputted signals 88.2K times a second (44.1K times per channel). At 20 kHz, the left and right signals had about 90 Degrees phase error relative to each other.

Sound is 1080 feet per second, about 1 foot per millisecond, which is about 2 cycles of 20 kHz per inch. 90 Degrees is 1/4 cycle, or about 1/8 inch at 20 kHz.

A set of left and right loudspeakers are positioned exactly the same distance from the listener. The listener turns his head 1/8 inch total (90 Degrees at 20 kHz). Once the correct phase is established, the listener must not to turn the head again, not even 1/8 of an inch.

If you are worried about the relative phase versus frequency of just one channel's amp and loudspeaker drivers, be sure not to move your head up or down even a very small distance, you will disturb the relative phase of the woofer versus the tweeter.

How we seem to worry about the inner layers of the onion, when the outer layers have much larger problems.
 
Slight ringing can't result in ugly square wave unless there is no damping and/or F3 is much lower than Fres which would mean the amp might be unstable and/or the transformer is not really suited for this job.

For me, optimal behaviour is such that cut-off is just below resonance. Extending the FR beyond the resonance is pointless because the phase will rotate anyway and the primary load will show a big dip in module. No serious power can be had there and not good if fbk has to be used.

What is the source impedance in case 2? what is the primary load? Fres at 135 KHz is a rather good transformer by any standard.
 
1. If the amplifier Does have negative feedback, I would probably choose transformer #1.
You have to worry about square waves into resistive loads during a Reviewers testing (or a audio club member's testing). You also have to worry about the real world loading from loudspeakers and loudspeaker cables, when using the amp for listening to music.

Opinions are welcome for both cases - NFB available or not. Speaker and cable loading have a significance as well. I would probably strive to the optimal solution for music listening, compared to resistive load measurements. Cable capacitance is one evil and it's in parallel with the secondary inductance.


2. If the amplifier does Not have negative feedback, And is either a single ended triode amp or a push pull triode amp, I would use either transformer.
That is because the music sources I use do Not have frequencies higher than 20kHz. There will be No ringing in that application. And sense there is no negative feedback, you do not have to worry about stability.
So how about some other criteria for selection of the transformer:
Which transformer is less expensive?
Which transformer has more inductance?
Which transformer has better primary and secondary DCR?
Which transformer has more laminations?

Because I am not making a commercial amplifier, with some Reviewer or audio club member trying to criticize the amp's bench performance, I do not have to worry about some ringing at ultrasonic frequencies.

Both transformers will have the exact same core, number of winding, labour, hence price. There will be a slight alteration in insulation quantity and type, changing Cp vs Ls.

Slight ringing can't result in ugly square wave unless there is no damping and/or F3 is much lower than Fres which would mean the amp might be unstable and/or the transformer is not really suited for this job.

Of course, if F3 is much lower than resonance, this will be an unacceptable situation.

For me, optimal behaviour is such that cut-off is just below resonance. Extending the FR beyond the resonance is pointless because the phase will rotate anyway and the primary load will show a big dip in module. No serious power can be had there and not good if fbk has to be used.

In my case, FR beyond the resonance, either by Cp or Ls was just a bonus, not a search. I agree with you.

What is the source impedance in case 2? what is the primary load? Fres at 135 KHz is a rather good transformer by any standard.

It's a 3k/4R transformer, dedicated for tubes with a Rp similar to the one of 300B (650-800R)
 

45

Member
Joined 2008
Both transformers will have the exact same core, number of winding, labour, hence price. There will be a slight alteration in insulation quantity and type, changing Cp vs Ls.

In that case just go for the one that performs better overall.

It's a 3k/4R transformer, dedicated for tubes with a Rp similar to the one of 300B (650-800R)

In general you can keep the option 2 for another tube. For example if you run current production KT66 in pentode mode with 250V on G2 and 300-310V on the plate at 85 mA you will get at least 9W into 3K. So it's 300B territory. If you use 10% Schade fb then for gm around 5-6 mA/V the Zout will be 1.6-2 K....:D
In general, for well behaved FR you don't need exactly Q=0.7. Anything <1 should not show any ringing.
All this always assuming that SQRT(Ls/Cp) is less than Rl (3K in this case, so it's good).
 
If one were designing the OT for CFB driven by a pentode, then the cathode and plate OT windings would want to be optimized differently. Cathode being low Z would like low leakage L to secondary, and plate being high Z would like low distributed capacitance. Now you can optimize the OT performance far better when only one parameter needs to be minimized for each winding. One could optimize the cathode and plate winding Q parameters separately.
 
Last edited:

45

Member
Joined 2008
Hey Don, with CFb if one is not using large amounts the low Ls is almost trivial. For example, the last OT I am going to get soon will have 1/6 CFb ratio and its Ls will be over 200 times smaller than the full 5mH. So it should be around 20 uH.
I think that should low enough.
 
Should be good.

If the CFB is small enough, it could even be bifilar with the secondary, with the small %CFB also taken back to the driver for more loop gain. Providing it remains stable.

With the pentode plates acting as current sources, the plate winding portion may not even need interleaving. Just fully minimize the distributed C. Just a little extra B+ headroom for the plate "current sources" to overcome the leakage L then, which doesn't really require any B+ change. No one wants full power at the HF end anyway. This could make for a cheap to wind OT, with reasonably high performance.
 
This could make for a cheap to wind OT, with reasonably high performance.

Indeed.
I have a design for a small OT just big enough to get 18W @20Hz without saturation which I will get done in August when I' ll go back to Italy for holidays. 1/6 CBf with the 10GK6 in PP. Max Pout optimised for 5R (where total primary impedance is the standard 8K) basically 40:1 step down, including CFb. Core size 2.5x4 cm. Total Cp for the plates should be no more than 0.3 nF according to my estimate.
 
Opinions are welcome for both cases - NFB available or not. Speaker and cable loading have a significance as well. I would probably strive to the optimal solution for music listening, compared to resistive load measurements. Cable capacitance is one evil and it's in parallel with the secondary inductance.

Answer: Of course, I strive for optimal performance for music (and therefore with loudspeaker loads). Testing the amp with load resistors is great, it tells me lots about what I did right and/or wrong. THEN I connect the amp to loudspeakers. Note: I would only use Negative Feedback for Ultra Linear, and Pentode modes (for either single ended or push pull).

I do Not use negative feedback for Triode, Triode wired Pentode, and Triode wired Beam Power output tubes. I design what primary impedance to use, so that "Triode" rp gives reasonable damping factor and reasonable distortion (but tradeoff power).

Both transformers will have the exact same core, number of winding, labour, hence price. There will be a slight alteration in insulation quantity and type, changing Cp vs Ls.

Answer: If turns and core are the same, and only the insulation changes (i.e. dielectric constant, and thickness) how did you get:
Ls = 0.69mH (transformer #1)
and
Ls = 1.13mH (transformer #2)

You must have used thicker insulation on the transformer that has Ls = 1.13mH. That would give slightly less coupling from winding layer to winding layer. Right?
 
They both seem like good transformers.

Isn't Q dependent on L,C, and Rs? If the windings are the same, just different insulation, how can the Q (and thus Rs) be so different?

Anyways, for a class-AB PP design I understand you want the lowest possible leakage inductance so that the tube shutting off in the class-B portion of the signal doesn't produce horrible spikes. Leakage inductance is worse than leakage capacitance in most cases I'd think. Unless we're talking very high impedances (low current high voltage which case capacitance is worse).

My vote goes to #1.
 
SemperFi,

Yes. Not just the factors that I said in post # 11. You also caught what I did not think about . . . the possibility of the different Rs. In order for Rs to be different if the number of turns are the same, the wire size has to be different. But different wire sizes will change the capacitance too, smaller wire size will pack closer, but have less surface area from top to bottom that is next to the other surface area from top to bottom. It will also change the number of turns across the full width winding, since smaller wire size will pack tighter. That will change the number of layers required to get a specific inductance (total primary inductance, I do not mean Ls).
 
Last edited:

PRR

Member
Joined 2003
Paid Member
> different wire sizes will change the capacitance too

Not much. Each turn counts for less, but you typically put on more turns. You can usually estimate C simply by the area of the winding, not the wire size.

That's for audio. If you are working high frequencies (and no bass) or high voltages, you wind with appreciable space between conductors (perhaps by laying a thread between turns). If only half your area is copper, C is about half.
 
Answer: If turns and core are the same, and only the insulation changes (i.e. dielectric constant, and thickness) how did you get:
Ls = 0.69mH (transformer #1)
and
Ls = 1.13mH (transformer #2)

You must have used thicker insulation on the transformer that has Ls = 1.13mH. That would give slightly less coupling from winding layer to winding layer. Right?

You answered to the question yourself. Yes, by doubling the distance between the layers, you decrease the capacitance by half, but Ls doesn't double. You can find an optimum if you play with the insulation parameter. Second comes the dielectric constant. I have some experiments in mind considering a hybrid air-nomex insulator.


They both seem like good transformers.

Isn't Q dependent on L,C, and Rs? If the windings are the same, just different insulation, how can the Q (and thus Rs) be so different?

The "problematic Q" is a series RLC circuit and is dependent of Ls (leakage), overall capacitance (Cp) and Rs. Different insulation (type and thickness) alter Ls due to coil thickness increase/decrease and Cp due to distance and/or dielectric constant change).

There is a resonance before this one, dependent of Cp and the main inductance L being parallel to it, but its Q is naturally so low that this resonance is practically insignificant and can be hard to measure. Its nature is more problematic in inductors with a low main inductance, like air coils for speakers.
 
By the way, I have done once the output transformers for a 60W Dynaco Mark III kit (DIY clone) with KT88's and they had some 3 dB peak at the resonance around 77 KHz. This was really easy to deal with. In fact the original Dynaco design already had a small 12 pF between B+ end of the anode resistor and grid of the phase splitter and another 390 pF cap in fdbk loop. Both these caps are there to reduce gain above 40 KHz and get the correct phase. Just needed to adjust those values.

The EI core transformers had 77KHz resonance due to 5.3 mH Ls and 0.8 Cp. As the UL output stage with KT88 has a Zout of about 6.5K and the Zaa is 4.5K and it had to be affordable there wasn't much choice. But it was as good as it gets in the audio range. In fact it could deal with 60W @25Hz with just 0.8T induction and over 400H primary inductance for very low distortion.
 
50AE,

I used to have access at work (after hours) to a German $60,000 4GHz Vector Network Analyzer (VNA) that could make precision measurements as low 10Hz. It used 'external' mode, and a T connector so it could get down to 10Hz (classical Return Loss Bridges need not Apply).

The very high dynamic range, external mode, T connector, and precision cal kit (open, short, load) with high order math in the software, gave it the capability to make very good tests on Output transformers, and Interstage transformers (as well as on amplifiers).

The many output transformers I tested had primary resonances (L and Distributed-C) that were between 500Hz - 2kHz. These were high impedance at resonance, and yes as you noted are swamped out by the drive impedance (especially by triode output tubes).

Swamping this resonance out with an un-fedback Pentode or Beam Power tube does not happen, the only swamping that occurs is from the secondary load (and local or global negative feedback).

The vector network analyzer, plus some external resistors for driving the primary, and for loading the secondary allowed the following tests: Bandwidth I/O Phase Power Loss Primary Inductance Primary Distributed C Leakage Reactance Primary to Secondary C Damping Factor at the secondary versus Frequency (When the primary was shunted with a resistor that simulates the driving plate impedance).

I also tested amplifier's frequency response, gain, and phase versus frequency. I used load resistors, and also loudspeaker loads, and compared the results. The amps tested had triode output stages, and did not have negative feedback.

Perhaps there were other tests that I forgot that I did with that VNA years ago.

I miss having access to that VNA.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.