1:1 quadrifilar wiring
- Thread starter dogears
- Start date
Help Support GroupDIY Audio Forum:
Winston OBoogie
Well-known member
series : series.
series - parallel. 
what is in and out impedance that you are trying to match?
how much current is involved?
parallel will halve DCR and therefore be less lossy current wise.
what transformer?
API 2503? Quad 8?
since the wire is all wound side by side, there is no need to worry about coupling, that is, in some situations, you might have four winds stacked up on each other, so you might wanna balance the leakage and DCR by using 1 and 3 pri then 2 and 4 sec.
usually a quad wound has low inductance due to few turns, so if it is bass you are after, a series connection would double the turns and therefore square the inductance, so your 250 mH is now 1 Henry.
20 hz at 1 Henry = 6.28 x 20 x 1 = 1250 ohms of Inductive Reactance. so if you have a 150 ohm source, it will degrade bass less.
If maximum power transformation is what you are after, you try to match the xfmr to the load so 250 mH would = 20 x .25 x 6.28, or out first answer divided by four which is 312.5 ohms. This would lower losses if you have a lot of current you want to pass thru the coil.
Only thing left to consider is hi freq response. There is a lot of capacitance on these quad wound jobs. Sometimes the wire is twisted to provide a smooth transistion in frequency response as the core drops out at around 1500 Hz and the air core properties of the coil take over. Air has amazing frequency response flat from 0 Hz to Infinity. Therotically (sp) speaking. So if you can get the Stray C right, you can go linear all the way up to 1 mega Hertz.
Problem is, which hookup is gonna do this? Series or Parallel?
That my friend is easier to determine by 5 minutes with the scope than 10 million equations involving phase shift and J operators to determine the complex impedance of the reactive elements.
I would intuitively think that the series connection would be better for hi end. Since i am usually wrong, that means that parallel would be better.
Now, what if we wire it in series and run it out of phase that is, have the polarity of the signal flipped as it passes thru the xfmr? Now you have AC voltage gradients scattered throughout the coil such that the difference in potential along the length of the parallel wires will contribute to who knows what! I ain't no genius i jus rip em apart. ;D
man, i need to vcacuum this keyboad, i had to d 10 edits, isnt ther a privw function?
what is in and out impedance that you are trying to match?
how much current is involved?
parallel will halve DCR and therefore be less lossy current wise.
what transformer?
API 2503? Quad 8?
since the wire is all wound side by side, there is no need to worry about coupling, that is, in some situations, you might have four winds stacked up on each other, so you might wanna balance the leakage and DCR by using 1 and 3 pri then 2 and 4 sec.
usually a quad wound has low inductance due to few turns, so if it is bass you are after, a series connection would double the turns and therefore square the inductance, so your 250 mH is now 1 Henry.
20 hz at 1 Henry = 6.28 x 20 x 1 = 1250 ohms of Inductive Reactance. so if you have a 150 ohm source, it will degrade bass less.
If maximum power transformation is what you are after, you try to match the xfmr to the load so 250 mH would = 20 x .25 x 6.28, or out first answer divided by four which is 312.5 ohms. This would lower losses if you have a lot of current you want to pass thru the coil.
Only thing left to consider is hi freq response. There is a lot of capacitance on these quad wound jobs. Sometimes the wire is twisted to provide a smooth transistion in frequency response as the core drops out at around 1500 Hz and the air core properties of the coil take over. Air has amazing frequency response flat from 0 Hz to Infinity. Therotically (sp) speaking. So if you can get the Stray C right, you can go linear all the way up to 1 mega Hertz.
Problem is, which hookup is gonna do this? Series or Parallel?
That my friend is easier to determine by 5 minutes with the scope than 10 million equations involving phase shift and J operators to determine the complex impedance of the reactive elements.
I would intuitively think that the series connection would be better for hi end. Since i am usually wrong, that means that parallel would be better.
Now, what if we wire it in series and run it out of phase that is, have the polarity of the signal flipped as it passes thru the xfmr? Now you have AC voltage gradients scattered throughout the coil such that the difference in potential along the length of the parallel wires will contribute to who knows what! I ain't no genius i jus rip em apart. ;D
man, i need to vcacuum this keyboad, i had to d 10 edits, isnt ther a privw function?
Winston OBoogie
Well-known member
trobbins said:
Interesting. Checking relevant quad-filar transformers from Jensen, Cinemag, Sowter, and Lundahl - all show a series connection in the online published data sheets and test circuits. I can't see any benefit to parallel connection myself but, each to each I guess
if you wanted to limit the 1 MHZ response you could ground one or two windings, forgot about that scenario.
you could also leave two windings floating and see what happens,
and you can always use he unused wind for a VU meter, noting that this could damage the signal if not buffered, however, usually the API OPT is pretty stiff as far as driving capacitive loads,
you could also use one of the winds as a NFB loop, put it into a resistive pad, phase linearty is very good in the quad wound xfmr.
you could also wire the unused wind into a pot or rotary switch equipped with various capacitors in order to enact a tone control.
you could also leave two windings floating and see what happens,
and you can always use he unused wind for a VU meter, noting that this could damage the signal if not buffered, however, usually the API OPT is pretty stiff as far as driving capacitive loads,
you could also use one of the winds as a NFB loop, put it into a resistive pad, phase linearty is very good in the quad wound xfmr.
you could also wire the unused wind into a pot or rotary switch equipped with various capacitors in order to enact a tone control.
Thanks CJ. Awesome brain food.
As for leaving them floating - I've found that can do odd things to transient response. Folks at Jensen recommended loading the unused winding (if 1:2) with a 10k or something rather than leave it dangling.
I'd always just done series:series, assuming mo is betta, but I'd never really given it serious thought.
As for leaving them floating - I've found that can do odd things to transient response. Folks at Jensen recommended loading the unused winding (if 1:2) with a 10k or something rather than leave it dangling.
I'd always just done series:series, assuming mo is betta, but I'd never really given it serious thought.
+1to what CJ says.dogears said:
It depends very much on the surrounding topology.
I had custom xfmrs made for me that I used in several different ways. As 600:600 iso they were series:series, but as mic splitters they were 150:150:150:150; then as output xfmrs, one winding was the drive, another the NFB, and the two remaining were in parallels for the output.
Most standard quadfilars are optimized for 600:600 iso because that's where the demand is, so connected series:series, but their use as 1:2 or 1:3 is usually well documented.
john12ax7
Well-known member
dogears said:
I've wondered what is optimum for 1:2. Is it preferable to leave a winding unused (terminated), or would it be better to parallel the primaries?
Parallel the pris. Minimizing DCr of the resultant pri reduces THD and improves LF response.john12ax7 said:
john12ax7
Well-known member
abbey road d enfer said:
Great, thank you.
I spent some time experimenting and simulating xfmrs, using detailed models that included multiple stray capacitance. The best case is a balanced source and a balanced receiver. When leakage inductance and stray capacitance start to roll-off, capacitive coupling between windings takeover and extend HF response way beyond the AP's limits.CJ said:
Worst case is when the primary is grounded on one side and the secondary drives a receiver that's grounded on the other side; then capacitive coupling is out-of-phase with magnetic coupling, resulting in significant HF roll-off. Simulating with lumped elements results in severe notches in the response. In practice they don't exist so much; using a model with continuous transmission lines gives a more accurate answer.
Winston OBoogie
Well-known member
abbey road d enfer said:
Lundahl are the only company I've seen that have sometimes given response measurements for grounded and ungrounded windings, the results, of course, tallying with your own modelled results. ***
I will say that I've never seen an example of a quadfilar transformer being used as a 150 ohm mic splitter as you did with your own custom transformers. Generally, these are seen with an electrostatic shield between the windings but, hey, if it works
*** Edit: on the whole, manufacturer's data is mostly thin on the ground in my opinion. Of the bigger known names, Carnhill, followed by Sowter seem to be the worse in that regard.
You're absolutely correct. They also have a magnetic shielded enclosure, which IMO is even more important than electrostatic shielding. However I just tried it, put that in a box and ended up selling 1000's of them... people just learned not to put them close to guitar amps, keyboards and equipment racks.Winston O'Boogie said:I will say that I've never seen an example of a quadfilar transformer being used as a 150 ohm mic splitter as you did with your own custom transformers. Generally, these are seen with an electrostatic shield between the windings but, hey, if it works
There are a couple of threads here about this. PRR, whom I respect a lot, was of the opinion that the best spec is describing a preferred usage. That is one of the few points where we may disagree somewhat. Anyway, the level of variability of primary characteristics, particularly inductance implies lengthy specs that many potential buyers would be deterred by.
[/quote]
Winston OBoogie
Well-known member
abbey road d enfer said:
Indeed, that would be absolutely vital here. I'd just assumed and was envisioning yours as being potted in a large shielding can, but if an enclosure worked and you sold 1000's , then your design was good and that's that. Nice one. I suppose mid/side encode/decode would have been another use for these units.
abbey road d enfer said:
Maybe it's a result of an increased market to the DIY crowd? But it frustrates me when a spec merely says for instance "push-pull plates to line" and the turns ratio. I want/need to know a whole lot more info before spending £100+ on a transformer. A Sowter can cost that much and sometimes not even have a data sheet of any sort.
It could have been a possibility, but it was not a market I was addressing.Winston O'Boogie said:
That's right, but you can have a nice conversation with Brian S, and then you will know much more. You have to convince him you know what you're talking about.
Rob Flinn
Well-known member
abbey road d enfer said:
Is there any case where driving to ground through a resistor would improve this scenario ? Rather like the quasi balanced outputs one gets nowadays.
No. Adding a resistor in the drive path is detrimental to THD and frequency response.Rob Flinn said:
Driving symmetrically and loading the secondary with a balanced receiver is perfect, though.
Contrary to common belief and some mfgrs litt, driving a xfmlr from a balanced driver such as THAT 16xx, DRV134/5, SSM2142, is not a good solution, since the output impedance is high and generates THD. There are topologies that introduce a resistance in series with the primary, used to sense current, and fed back to the drive circuit in order to implement negative output resistance. Optimum THD and LF response is obtained when negative resistance compensates (almost) exactly the primary resistance.
Winston OBoogie
Well-known member
Here's the Lundahl pdf showing the basic idea of sensing the current to create this "almost" inverse of the primary dcr. In practice, the sense R is best adjusted in situ for lowest THD but, If in doubt, err on the low side.
Indeed: too much negative impedance turns the circuit into an oscillator.Winston O'Boogie said:
