Transformer phono correction

the road not taken.
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Transformer phono correction

Post by piotr_grzesik »

I've been at phono stuff recently and was reading that Audio Note are using Transformer RIAA correction.

CV pointed me in the direction of this (I assume the Audio Note solution is similar to this):

http://www.aml.nm.ru/RX_corrector_eng.htm

Looks like an interesting way of doing things. Does anyone have any comments? Dave, how hard to you reckon it would be to make such transformers?
Too much stuff and not enough time!
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey pete,

attached is the PDF of the corrector. I have thought about this for some time now and the only tough thing is getting the two dissimilar gaps. I have a few ideas how this could be done but i suspect it will take a fair bit of trial and error.

Last I heard ANUK has been playing with the concept but have not used it in any production products.

dave
Attachments
RX_corrector_eng.pdf
(264.9 KiB) Downloaded 1100 times
DowdyLama
Posts: 104
Joined: Thu Jan 27, 2005 5:42 pm
Location: Blue Ridge Mtns, NC
Contact:

Post by DowdyLama »

This is pretty fascinating - thanks for the post Pete!

I'd be very interested in experimenting with this approach...especially involving some of the tubes I really like 8)
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Ah, I figured you would already be onto it Dave! :D

These were the pages which set me off, which seem to indicate a production piece:

http://www.audionote.co.uk/news/article ... 6_m9.shtml

http://www.audionote.co.uk/products/pre ... o_01.shtml

I'm sure win true Audio Note style it's mind-bendingly expensive. :shock:

I'd also be up for some guinea pig work, if you're inclined!

What tubes were you thinking of, Jim? 450TL or such? :lol:
Too much stuff and not enough time!
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Hey guys,

I just pointed another person who was asking about this here so we can do the work together. If we come up with a spice document that works then I can start on the transformers. I have too much on my plate to dive into this project in spice, However coming up with a transformer that matches the behavior of the models in spice is something I might enjoy every night as I curse your names.

I think the original author came up with a very clever idea and I'll insist he gets full credit and I believe it may have been CV (shinebox) who pointed me originally to this concept 4 or so years ago. I know it was mentioned on ROMY's site so maybe that is where my knowledge of the concept came from.

shines do you recall?

dave
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

Post by shinebox »

Darned if I remember... I think the author was an influence on the Cat in his formative years... I suspect it may have been at GSC where I saw it too.

I haven't actually read the article in years so I must revisit. My understanding of leakage inductance was that it was as much a function of the sectioning geometry as the air gap, though possibly the latter dominates? Would be easier to control the former (pun intended) I guess.

I assume you have a ton of references Dave; I can dig out my old power electronics book which has some formulae if you like?

cheers
shines
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

what are you still doing up? Does that machine have you learning yet?

From my meek perspective, it is the two dissimilar flux loops that contribute 100% to the solution of the leakage. I do not see any winding geometry having any effect in the ranges involved... unless you go air cored with three coils :-)

didn't he mention something like 20hy's of leakage? that isn't going to happen with geometry. The coils on the outer legs of the EI and the gap on the center leg creates two different flux linkage paths between the windings and is the solution I respect.

I know I have seen the picture of the EI core with the windings on the outer legs somewhere in the old texts. The goal was to take advantage of the leakage from the dissimilar flux paths but the concept was useless until i saw the RX phonocorrector and the connection was made.

dave
Nickel Core
Posts: 14
Joined: Wed Sep 26, 2007 8:52 pm

RX corrector

Post by Nickel Core »

Hi There,

I was the guy who talked to Dave about the RX corrector before. There is a lot to tell, but here are some points:
  • The RX corrector was first proposed by Anotoly Likhnitsky ithe Russian 'Audiomagazine' here: http://www.aml.nm.ru/RX_corrector_eng.htm;

    Likhnitsky worked with Sergey Sokolov. In September 2005 a Patent was granted for the Rx theory (attached) and in October 2005 a Patent was granted for the physical structure of the transformer (attached);

    The RX corrector can be realized in several ways, with two or three inductors;

    Peter Qvortrup of ANUK pointed me to the work of Likhnitsky;

    Audio Note UK never implemented a commercial product using the RX corrector (also the M9 as no RX corrector);

    Audio Note was/is planning (as I understood) to implement the RX corrector in the upcoming M-10 Phono, I do not know what the current status is;

    I tried to model the RX corrector in LTSpice, but did not yet get it to work. I made a spreadsheet to calculate all the value's, but I don't think I have it correct yet. Further LTSpice does not let you implement the desired coupling (you need something like K1 L1 L2 0.8xxx and K2 L1 L3 0.98xxx)), this is not a allowed physical construction in LTSpice);

    You can model the Rx corrector with 2 separate transformers K1 L1 L2 and K2 L3 L4), but with this I'm unable to come up with proper results yet;

    You need to realize that by using the leakage inductance to introduce the correct slope, you are actually replacing a well known component as a capacitor by a fairly 'uncontrollable' and 'virtual' component which may lead to all kind of problems.
But I LOVE to try it, it somehow appeals to me. If I can build a two stage amp where the driver transformer does the RIAA correction, then that would be SOME THING!.

Regards,
Nickel Core
Attachments
RxTransformer1.pdf
RX-Theory Patent
(482.47 KiB) Downloaded 528 times
RxTransformer2.pdf
RX-Physical construction Patent
(1.17 MiB) Downloaded 571 times
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Thanks guys!

NC, can we not simply model it as a perfect transformers + leakage inductances?

I did a couple - RX01 is just a model using voltage sources to model the transformers and RX03 is an attempt at a practical circuit using 437A. I;m not sure whether these are complete and we need to account for anything else? Stray capacitances?

It's pretty sensitive to values, which is maybe a reason for lack of use in a production piece.

What do you guys think?
Attachments
RX03.jpg
RX03.jpg (126.78 KiB) Viewed 34047 times
RX01.jpg
RX01.jpg (237.59 KiB) Viewed 34047 times
Too much stuff and not enough time!
Nickel Core
Posts: 14
Joined: Wed Sep 26, 2007 8:52 pm

Post by Nickel Core »

Hi Pete!

Very nice!

Yes, to model it as a perfect transformer + leakage inductance was my first idea, but I wanted to 'fine tune' the coupling coefficient to translate back to the gap size for the transformer (as per Likhnitsky's formula's). Am not very good in math, so I was unable to translate the leakage inductance back to the gap size .

How did you arrive at the current values? Did you calculate them or did you tune by trial and error in LTSpice? I also see that that you a 1:1:1 inductance ratio between primary and secondary/tertiary and that you added series resistance to the secondary and teriary inductors (additional since both inductors are 100H).

Could you post the .asc file? I modeled your proposal, but I don't get quite the same results...

EDIT: the slight difference in result might be caused by slightly different models for the tube. This got me thinking that the design is very sensitive to tube variances and aging.

This all might be too much for a commercial product but for DIYers like us ;-)

Regards,
NC
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Hey Pete and NC,

Pete, your model looks like a typical LRiaa which is very source impedance dependent. Essentially the Rp of the tube defines the behavior of the filter. My solution to this was to simply place an adjustable CCS (decoupled) in the plate circuit to adjust the Rp of the tube to tune the filter and account for tube variances.

I'm not sure I see the need to keep the leakage inductances within the transformers when a simple external series choke will do a similar thing and allow for ease in tuning. I guess I'm wondering what is gained?

dave
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Hi guys,

I'm not and expert on modelling, so I don't know if that's the correct model for this. The text talks about the time constants coming from the leakage inductance which itself relies on lossy coupling. I'm not sure how the leakage inductance and 'k' factor should be considered in SPICE. My question would be - is the leakage inductance simply derived from the 'k' factor by SPICE and then it's correct to simply model as I've done? Or not? I'm not sure! It's certainly not possible, as NC says, to define two different couplings to different secondaries...

But Dave, it does seem somewhat pointless when you put it like that, because yeah, essentially that's what it seem to be. :idea:

NC, ASC files attached for you to play with.
Attachments
RX_1.asc
(2.79 KiB) Downloaded 663 times
RX_3.asc
(3.48 KiB) Downloaded 563 times
Too much stuff and not enough time!
Nickel Core
Posts: 14
Joined: Wed Sep 26, 2007 8:52 pm

Post by Nickel Core »

Hi Guys,

Well, the only LRiaa design I know of (from Steve B) uses two IT's in series. This design is simpler than that which I like.

Why the leakage inductance should be in the transformer is a good question. The way I see it is that this 'virtual' component might well be a 'perfect' one as long as you get the value right. Leakage inductance has no copper losses, no stray inductance etc. It is there, and it is there for free.

Obtaining large leakage inductance might well prove a just a big hassle as obtaining very low leakage inductance, but it can be done as Sokolov showed. I also don't see (yet) if and why a RX transformer would be less prone to source impedance variances than a LRiaa. This should be investigated further because IF the design is less sensitive to source impedance, than the RX transformer has a big advantage (as to avoid a CCS, positive current feedback or more (step-down) interstage s).

At the end, the proof is in the eating. I might just 'sound' better even if there is no engineering explanation.

Pete, the different result I got was because of the series resistance of L1 (500R) which didn't show up in the schematic.

Anyway, much to think about...

Regards,
NC
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

I believe the K factor defines leakage by the simple ratio. A 100hy 1:1 with a K of .99 will have 1hy of leakage inductance. I confirmed this empirically by looking at the high frequency response of a 1:1 with such a factor and by changing the K to 1 and placing an uncoupled 1hy choke in series.

I am trying to come to grips with pete's sims. the coupled transformer secondaries essentially put the filters in parallel, yet externally they are connected in series.

dave
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Nickel Core wrote:Hi Guys,

Well, the only LRiaa design I know of (from Steve B) uses two IT's in series. This design is simpler than that which I like.

Why the leakage inductance should be in the transformer is a good question. The way I see it is that this 'virtual' component might well be a 'perfect' one as long as you get the value right. Leakage inductance has no copper losses, no stray inductance etc. It is there, and it is there for free.

Obtaining large leakage inductance might well prove a just a big hassle as obtaining very low leakage inductance, but it can be done as Sokolov showed. I also don't see (yet) if and why a RX transformer would be less prone to source impedance variances than a LRiaa. This should be investigated further because IF the design is less sensitive to source impedance, than the RX transformer has a big advantage (as to avoid a CCS, positive current feedback or more (step-down) interstage s).
I'm thinking whether we can go the pentode/cascode route to set the source impedance, i.e. use a plate resistor in parallel with the transformer. Possibly it doesn't work here as it may not be as constant as with the LCR circuit. Need to think more/have a play.
At the end, the proof is in the eating. I might just 'sound' better even if there is no engineering explanation.

Pete, the different result I got was because of the series resistance of L1 (500R) which didn't show up in the schematic.

Regards,
NC
Aha!, yes in the first schematic I didn't separate them out. In the second one I did, because it's easier to see that way, and as we're modelling using "perfect" components. Might be sensible to add capacitances as well? Though I have no idea what figures to use... over to Dave!
:D

dave slagle wrote:I believe the K factor defines leakage by the simple ratio. A 100hy 1:1 with a K of .99 will have 1hy of leakage inductance. I confirmed this empirically by looking at the high frequency response of a 1:1 with such a factor and by changing the K to 1 and placing an uncoupled 1hy choke in series.
That makes sense, nice to have it confirmed. Do you know if the k-factor derives any other parameters? Or is it just the inductance?
I am trying to come to grips with pete's sims. the coupled transformer secondaries essentially put the filters in parallel, yet externally they are connected in series.

dave
Hmmm...perhaps it makes sense to use two transformers? To make the job easier? But then it is *very* similar to the LR....
Too much stuff and not enough time!
reVintage
Posts: 365
Joined: Fri Sep 07, 2007 9:54 am
Location: Sweden

Post by reVintage »

Interesting concept!

The patent holds some totally unrealstic figures though. Somewhere is said that N2/N1 should equal 11,8. This means that if the secondary N2 should be set to 800H it also means primary should be ca 6H :roll: ! I am to lazy to go through the maths and adopt it to LTSpice, but all answer are in the patent!

I have anyway taken the liberty to adjust the RX3 *.asc to give you better tools to work with. So in this one the values are theoretically correct(incl. Neuman 4th) from signal to primary. Good luck!
Attachments
RX3LS.asc
(4.42 KiB) Downloaded 603 times
Brgds
Lars
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Thanks Lars!
Too much stuff and not enough time!
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey guys.

Lars, I ran your version and the pic is below. When I put a 4700uf bypass across Pete's Rx3 I saw a similar response.

Pete, what is the rational between the progression between RX1 and RX3? what purpose do the extra parts in Rx3 serve?

Finally I am somewhat concerned about the load put on the tube at low frequencies (plot below).

Further concerns from the transformer perspective is getting a 1:1+1 transformer to behave properly particularly one that carried DC and has 100hy's of inductance per winding (I suspect trifilar will be problematic but have a few ideas)

The good news is i see 100hy's as way too much inductance. at no point do you ever want anywhere near flat response to 2hz in phono. If you stack several 6dB slopes just below 20hz, you have just made an effective rumble filter.

I'm just tossing ideas out there to further the discussion. Sorry if this seems scrambled, I'm just taking a break from some filed coil sims :-)

finally, damn spice for not allowing for various coupling factors between windings on a common core.


dave
Attachments
Assuming a tube with a 1K5 Rp we are loading things pretty heavy at low frequencies and gets lighter as you go up.  The odd thing is there is no loss in amplitude from the heavy load??
Assuming a tube with a 1K5 Rp we are loading things pretty heavy at low frequencies and gets lighter as you go up. The odd thing is there is no loss in amplitude from the heavy load??
Picture 3.png (19.61 KiB) Viewed 34068 times
Adding a bypass cap to the 35 ohm cathode resistor gives something closer to what lars has above.  If you sweep your B+ you easily see the variability the source impedance has.
Adding a bypass cap to the 35 ohm cathode resistor gives something closer to what lars has above. If you sweep your B+ you easily see the variability the source impedance has.
Picture 2.png (48.03 KiB) Viewed 34068 times
I noticed you dropped the K to .9997 and using a diode leaves you with a well bypassed cathode.
I noticed you dropped the K to .9997 and using a diode leaves you with a well bypassed cathode.
Picture 1.png (48.3 KiB) Viewed 34068 times
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Hiya Dave,
dave slagle wrote:hey guys.
Lars, I ran your version and the pic is below. When I put a 4700uf bypass across Pete's Rx3 I saw a similar response.

Pete, what is the rational between the progression between RX1 and RX3? what purpose do the extra parts in Rx3 serve?
RX1 was just a first step using the inverse RIAA to drive the "transformer", so was really just an experiment to see if I could get some values which looked ok. RX3 just added a real tube driving it. The extra components are simply the DCR of the transformer widings broken out so that we cans see them easily (using the 'perfect components' model).
Finally I am somewhat concerned about the load put on the tube at low frequencies (plot below).
I wondered if that was the case, and was also theorising that this is why the LCRs are good - they offer a more constant load. BTW, how do you get the impedance out of SPICE?
Further concerns from the transformer perspective is getting a 1:1+1 transformer to behave properly particularly one that carried DC and has 100hy's of inductance per winding (I suspect trifilar will be problematic but have a few ideas)

The good news is i see 100hy's as way too much inductance. at no point do you ever want anywhere near flat response to 2hz in phono. If you stack several 6dB slopes just below 20hz, you have just made an effective rumble filter.
Could this be a case of doing too much with one device? I this is where you knowledge and experience is key.
I'm just tossing ideas out there to further the discussion. Sorry if this seems scrambled, I'm just taking a break from some filed coil sims :-)

finally, damn spice for not allowing for various coupling factors between windings on a common core.


dave
Too much stuff and not enough time!
docali
Posts: 58
Joined: Thu Oct 16, 2008 7:16 am

Post by docali »

Hi,

using the attached circuit ist better and simpler playing with inverse RIAA.

Laplace=((1+s*3180e-6)*(1+s*75e-6))/((1+s*318e-6)*(1+s*3.18e-6))

This formula includes the "missing" 4th time constant. If you want to use the traditional transfer function just delete the fourth term (1+s*3.18e-6).
Attachments
RIAA_INV.asc
Inverse RIAA
(839 Bytes) Downloaded 538 times
reVintage
Posts: 365
Joined: Fri Sep 07, 2007 9:54 am
Location: Sweden

Post by reVintage »

Docali, if you run my "inverse RIAA" you see it already does this, including the 4th. But yours is not gain-adjusted so 5mV in gives ca 60mV out.

Please adjust gain by dividing by 12 and it will be as good the already published :wink: !
Last edited by reVintage on Mon Dec 22, 2008 2:43 pm, edited 1 time in total.
Brgds
Lars
reVintage
Posts: 365
Joined: Fri Sep 07, 2007 9:54 am
Location: Sweden

Post by reVintage »

Hi guys, the big problem with the present component values, will as Dave says be the extreme loading on the tube.

I suggest going back to the patent and read "between the lines" :wink: ! Personally I don´t get it totally :oops: !
Brgds
Lars
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Well Lars, we could always take your 1200 ohm LCR trick :-)

I doubled all the numbers and upped the DCR of the primary and got the following results. It works but the DCR numbers are a bit high for my tastes. Interesting thing is I only lose about a dB of overall gain???

Pete, To plot the load seen by the 437 simply plot the voltage at the plate and right click on the name at the top. This allows you to enter an expression and simply divide the voltage at the plate by the current through the transformer primary (L1) and you get your impedance plot. Click on the dB scale and convert it to linear or log to get ohmic values.

Finally I think I have the transformer for the LR version worked out... gotta do some tests tonight.

The 4800 ohms PRimary DCR bothers me, but we could just use an external 5K variable resistor to increase the load and tune the Riaa.


How do we add in the 4th constant on the LR?

dave
Attachments
RX_3_2X_load.asc
the file with the changes.
(3.69 KiB) Downloaded 545 times
a better load for the 437A  (note the expression at the top to get impedance)
a better load for the 437A (note the expression at the top to get impedance)
Picture 6.png (46.8 KiB) Viewed 34037 times
doubling of the load for the impedance plot below.  Note the bypass cap to lower the Rp of the 437.  the primary DCR is 4800 ohms and this bothers me but it still seems to work.
doubling of the load for the impedance plot below. Note the bypass cap to lower the Rp of the 437. the primary DCR is 4800 ohms and this bothers me but it still seems to work.
Picture 7.png (24.64 KiB) Viewed 34037 times
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

I also noticed that playing with capacitances on the inductors gave some interesting results... a few pf here or there made things ugly rather quickly.

dave
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Thanks Dave, that looks like an easier load. I wonder if it's possible to tweak the values to help that in the first case, but perhaps it is what it is...

I was kind of holding off modelling capacitances until we had the basis circuit, but given the sensitivity to source impedances, it doesn't really surprise me.

Speaking with CV earlier and he reckons that there's going to be coupling between the secondary inductors anyway, so it may be that this is even trickier.... :?
Too much stuff and not enough time!
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

I have found a lot of the issues spice suggests from inductor capacitances are really phantom and do not exist in reality. I also have to wonder how critical the source impedance really will be. I seem to recall when looking at steve's LRiaa in spice it too was very sensitive to source Z but in practice it seemed much more forgiving.

I wouldn't worry too much about the inductor coupling... since that can be dealt with by shielding if it turns out to be a problem.

dave
Nickel Core
Posts: 14
Joined: Wed Sep 26, 2007 8:52 pm

Post by Nickel Core »

Hi Guys,

I was just wondering, did you got the values by trial and error, or did you do the math?

Dave, the coupling cooficient of 1 seems unrealistic. Changing this will also give completely different results again. The sensitivity to values of the whole thing bothers me a bit.

Do we really need the 4th time constant?

NC
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

I'm fine with the K of 1. I trust the errors that show up with a K of less than 1 about as much as I trust the errors that show up abruptly with a change of 5pf of capacitance across a coil :-) I suspect this may be due to the total lack of damping spice applies. Maybe in reality losses come to our rescue.

As for the 4th time constant, Yes, I suspect we want it. If it was there in the cutting process, in order to undo the phase errors it created it needs to be in the playback chain. being down a dB at 20K doesn't bother me but the associated phase shift does.

dave
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

I think the 'k' factor is simply a shorthand way of expressing the leakage inductance (and perhaps other parameters) but in this case we're modelling them directly so we don't need to account for it elsewhere. I think it would be useful to added semi-realistic capacitances so that we're accurate as possible, especially in light of the sensitivity to changes in values.

It also seems to be sensitive to changes in load. :(
Too much stuff and not enough time!
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Hey Pete,

I is the "possibly other factors" that alarms me. Since the series inductance represents substantial leakage, why should a slight change in the K net such drastic resuts?

I'm also unclear how the K factor deals with frequency. At low frequencies I suspect the K is pretty close to 1 but above 20K it is substantially less due to the ineffectiveness of the core to couple. Since LT spice is designed for HF work I suspect the model assumes the coupling to be appropriate for the cores used (ferrites) and may not be entirely accurate for the materials we use.

Also as frequency goes up cores get lossy and losses will tend to de-Q any of the squirrelly things going on. you can put in a parallel resistance for losses, but this neglects the frequency dependent nature of core losses.


dave
Post Reply