My latest infatuation - transformers
Posted 25th October 2014 at 08:11 AM by abraxalito
Updated 6th November 2014 at 11:51 PM by abraxalito
Updated 6th November 2014 at 11:51 PM by abraxalito
I've found that some of the el-cheapo trafos (18rmb each) at a shop at the local electronics market are of split bobbin construction. This makes bodging up an audio OPT from two mains trafos a fairly straightforward matter.
I bought some with 9-0-9V and others with 0-12V secondaries. Then I disassembled them (fortunately they're not varnish dipped) and swapped out the 220V primary bobbin for the secondary of the other one. This gives me a trafo with 18V on the primary and 12V on the secondary, a step down of 1.5:1, impedance ratio of 2.25:1. So it makes a 4R drive unit appear as 9ohms to the chipamp.
And when I applied this to the output of the bass/mid of my chipamp (residing in the Phenix active speakers, its a TDA7265), apart from it sounding quieter I suddenly realized how much power supply noise I was still listening to. Incredible
So if you want to know if your chipamp PSU decoupling is really up to snuff, see how much difference a 1.5:1 trafo makes. Subjectively it 'distanced' the sound, Frank's 'disappearing speakers' trick once again
Update - bought some more 15VA trafos, this time with 15V secondaries. After taking them apart, I tore out the primaries (the wire is thin (0.15mm) so unwinding isn't recommended unless you have an hour or more to while away, just cut through the whole caboodle with wire cutters and recycle the wire as copper scrap). In lieu of the 220V primary I wound a 7.5V secondary (0.8mm wire, around 0.3R) making these 2:1 step down. The application is my now long-in-the-tooth D1080 MkIIs where a pair of TDA8947s drive the units. I'd recently discovered there was a substantial mis-match in that the units are 4R and the 8947 is not at all suited to such a low impedance load, given it has a bridged output. A 2:1 ratio trafo transforms these units into 16R, a perfect match. So far I've only done the trafo mod to the bass/mids but the effect is quite repeatable, much improved dynamics and a shifting backwards of the music so that the speakers now seem rather incidental to the goings-on. Don't forget a 3,300 or 4,700uF 'lytic in series with the primary though to prevent trafo saturation with the DC offsets typical from chipamps. The D1080s have undergone such a remarkable transformation that I shall have to buy up a few more pairs now....
Update2 - now have wound and installed the tweeter trafos. These are much easier and quicker to wind as the design equations indicate much fewer turns are required for HF duty. I use ferrite cores, similar to Ferroxcube EQ25 and 40turns for the primary, 20 for secondary. I'll post up a pic of the end result in due course. I've only wound a couple of pairs so far, the wire might benefit from optimizing as I'm concerned about HF losses and with the wire thickness I use (0.6-0.8mm) there may well be considerable proximity effect losses. Listening now and Leonard Cohen's voice has not been more engaging in its emotional impact, addictive stuff The soundstage depth is the best I've ever had from a pair of integrated electronics actives - these speakers have the active XO mods I talked about in an earlier post but still only use TL084s.
Update3 - I've now uploaded a pic of the tweeter trafo. As you can see its dminutive, under 30mm across and about 20mm high.
Update4 - I'm experimenting now with a smaller size of transformer. I found another vendor where their 10VA trafos cost 9rmb - half the price of the 15VA ones. But on disassembling the first one I found that they're not making the best use of the winding window - more copper can be got in These also aren't split bobbin which is an advantage for doing custom windings (rather than re-using ready made ones) because there's no shroud holding the two bobbins together, meaning more space for windings. So it may well be that a 10VA with optimized windings can equal or exceed a 15VA. Stay tuned
Update5 - added a pic of the 'unwrapped' electronics from my 2nd pair of D1080s - I figured I needed the electronics out in the open to better optimize the SQ. Besides that the cabinet was severely constraining how many caps I could tack on to the supply The trafos for bass/mid now are 9rmb 10VA types, wound for 2.5:1 step down. Needless to say the sound is awesome, a major incentive for me to persue higher step down ratios...
I bought some with 9-0-9V and others with 0-12V secondaries. Then I disassembled them (fortunately they're not varnish dipped) and swapped out the 220V primary bobbin for the secondary of the other one. This gives me a trafo with 18V on the primary and 12V on the secondary, a step down of 1.5:1, impedance ratio of 2.25:1. So it makes a 4R drive unit appear as 9ohms to the chipamp.
And when I applied this to the output of the bass/mid of my chipamp (residing in the Phenix active speakers, its a TDA7265), apart from it sounding quieter I suddenly realized how much power supply noise I was still listening to. Incredible
So if you want to know if your chipamp PSU decoupling is really up to snuff, see how much difference a 1.5:1 trafo makes. Subjectively it 'distanced' the sound, Frank's 'disappearing speakers' trick once again
Update - bought some more 15VA trafos, this time with 15V secondaries. After taking them apart, I tore out the primaries (the wire is thin (0.15mm) so unwinding isn't recommended unless you have an hour or more to while away, just cut through the whole caboodle with wire cutters and recycle the wire as copper scrap). In lieu of the 220V primary I wound a 7.5V secondary (0.8mm wire, around 0.3R) making these 2:1 step down. The application is my now long-in-the-tooth D1080 MkIIs where a pair of TDA8947s drive the units. I'd recently discovered there was a substantial mis-match in that the units are 4R and the 8947 is not at all suited to such a low impedance load, given it has a bridged output. A 2:1 ratio trafo transforms these units into 16R, a perfect match. So far I've only done the trafo mod to the bass/mids but the effect is quite repeatable, much improved dynamics and a shifting backwards of the music so that the speakers now seem rather incidental to the goings-on. Don't forget a 3,300 or 4,700uF 'lytic in series with the primary though to prevent trafo saturation with the DC offsets typical from chipamps. The D1080s have undergone such a remarkable transformation that I shall have to buy up a few more pairs now....
Update2 - now have wound and installed the tweeter trafos. These are much easier and quicker to wind as the design equations indicate much fewer turns are required for HF duty. I use ferrite cores, similar to Ferroxcube EQ25 and 40turns for the primary, 20 for secondary. I'll post up a pic of the end result in due course. I've only wound a couple of pairs so far, the wire might benefit from optimizing as I'm concerned about HF losses and with the wire thickness I use (0.6-0.8mm) there may well be considerable proximity effect losses. Listening now and Leonard Cohen's voice has not been more engaging in its emotional impact, addictive stuff
The soundstage depth is the best I've ever had from a pair of integrated electronics actives - these speakers have the active XO mods I talked about in an earlier post but still only use TL084s.Update3 - I've now uploaded a pic of the tweeter trafo. As you can see its dminutive, under 30mm across and about 20mm high.
Update4 - I'm experimenting now with a smaller size of transformer. I found another vendor where their 10VA trafos cost 9rmb - half the price of the 15VA ones. But on disassembling the first one I found that they're not making the best use of the winding window - more copper can be got in
These also aren't split bobbin which is an advantage for doing custom windings (rather than re-using ready made ones) because there's no shroud holding the two bobbins together, meaning more space for windings. So it may well be that a 10VA with optimized windings can equal or exceed a 15VA. Stay tuned Update5 - added a pic of the 'unwrapped' electronics from my 2nd pair of D1080s - I figured I needed the electronics out in the open to better optimize the SQ. Besides that the cabinet was severely constraining how many caps I could tack on to the supply
The trafos for bass/mid now are 9rmb 10VA types, wound for 2.5:1 step down. Needless to say the sound is awesome, a major incentive for me to persue higher step down ratios...Total Comments 101
Comments
-
The first observation I would draw attention to in comparison of sound of classD and classAB is my assumption for the sound quality differences of any amplifier is due to the modulation of the power supply by the audio signal. All discussion of how power supplies sound I base on that assumption until its overturned by evidence. So let's say that's the hypothesis I want to test - that SQ differences between amps are down to how the silicon part responds to the modulations on the power rails caused by the audio signal its outputting to the load.
With that assumption firmly in place, PSRR is important but so equally is the impedance of the power supply, as is the impedance of the load. These three aspects combined give rise to a FoM for the SQ of an amplifier. I start by assuming the PSRR is zero dB i.e. unity gain to power supply noise at the output.Posted 18th November 2016 at 11:12 PM by abraxalito
Updated 18th November 2016 at 11:17 PM by abraxalito -
what about statements like this "When your amp drives a purely inductive component like a transformer, it does not see your speaker wild reactive loadings but instead it sees a very stable high impedance load from by the inductance of the transformer which does not fluctuate under load or frequency"Quote:Originally Posted by abraxalito
1:1 has no benefit when connecting to speakers that I can see as speakers have no need of isolation, being already isolated totally.
if we had magically solved the psrr issue how would they compareQuote:Sound of classD vs classAB is a very interesting subject, one I've thought about a fair bit in the past couple of weeks. The reason for my thinking is playing with what's effectively an open-loop classD amplifier - my Popu digital amps - and adding lots of caps to its power supply.
PSRR of amps is a thorny subject because the measurements (if there are any at all) aren't normally done in a real-world scenario so tend to be next to useless. Why I like my Popu amps is they obviously have no PSRR to speak of, being open loop and classD. So they provide a fairly predictable basis for PSRR experiments, unlike other amps where the PSRR is practically unknown. Its a topic for a whole discussion rather than a comment on a blog post. You could open a thread on it but it would likely degenerate into a slanging match between those who listen and those who trust DSs.....
might be up for a small class d project sometime........Posted 19th November 2016 at 02:22 PM by nige2000
-
The first statement is utterly nonsensical - seems to totally misunderstand transformers. A transformer isn't 'a purely inductive component' (neither is any practical inductor) and that's just for a start.
As to your second question, I reckon classD would be better than classAB in that circumstance. ClassD's issues primarily stem from insufficient HF PSRR in my experience - but then my experience is rather limited to just a couple of examples of classD. For a small classD project, get hold of some cheapo TDA8932 boards, use your step-down output transformer with them and enjoyPosted 20th November 2016 at 01:03 AM by abraxalito
-
ok that might explain why i couldnt figure it outQuote:Originally Posted by abraxalito
The first statement is utterly nonsensical - seems to totally misunderstand transformers. A transformer isn't 'a purely inductive component' (neither is any practical inductor) and that's just for a start.
ill order some ill get it eventuallyQuote:As to your second question, I reckon classD would be better than classAB in that circumstance. ClassD's issues primarily stem from insufficient HF PSRR in my experience - but then my experience is rather limited to just a couple of examples of classD. For a small classD project, get hold of some cheapo TDA8932 boards, use your step-down output transformer with them and enjoy
can they be paralleled or stacked for impedance aswell?Posted 20th November 2016 at 05:30 PM by nige2000
-
Yeah I have built a 'stacked' TDA8932 arrangement, I have shared a picture somewhere on DIYA. But its flaw was using EI transformers so the HF was too rolled off due to leakage inductance. Doing it with toroidal transformers is the way forward and I'm thinking of a way to do it with only one transformer rather than several in stacked formation. I can't find a source of very small toroidal transformers, seems lowest size is about 30VA (other than those sealed up in fancy plastic enclosures, they go down to about 3VA) I think.Posted 21st November 2016 at 12:41 AM by abraxalito
-
hmm...
slow learning here
with signal is transformers all about step down... converting v to i and reducing the load/influence on sources.....Posted 30th November 2016 at 01:49 PM by nige2000
-
Transformers allow volts to be traded for amps - high currents are the enemy in poweramps, by running much higher voltage we're able to reduce the current for the same power output.
I've found a place locally that winds 15VA toroids so I'm going to order some for an amp project. Just have to decide what project I wanna do....Posted 30th November 2016 at 03:21 PM by abraxalito
-
at line level though, with the wimpy r2r source the only thing that makes sense is a major step down, to reduce the load on the source?
the trafos i got 10k/10k inserted between r2r out 625z and a 25k trim pot took a lot of energy out of the sound
assuming the r2r is now overloaded the only solution if i want to keep the trim pot is to buffer r2r into traffo into trimpot into amp?
is center taps any use in these trafos?Posted 30th November 2016 at 03:54 PM by nige2000
-
What I'd do for that 625R source impedance is buffer it using classA biassed emitter followers. Some examples of doing that on the 8 * TDA1387 thread. Then from those EFs go to a transformer 1:1. A step down here would lose volts and volts are needed for the next stage.Posted 30th November 2016 at 11:35 PM by abraxalito
-
so no ignoring the elephant then so eh..
have to rearrange some stuff then so
how would pass b1 type buffer fair out instead of ef's in such situations?
what voltages are required for the class a ef buffer?Posted 1st December 2016 at 12:07 AM by nige2000
-
EF buffer needs supply voltage about 1V more than the signal its handling in both directions. So suppose its handling 1VRMS that's 2.8V peak to peak, you'd need a 5V supply as minimum. You'd also need to control the DC offset tightly if you used the minimum supply voltage. Do you have a link to the pass b1 schematic, I'm not familiar with it?Posted 1st December 2016 at 12:41 AM by abraxalito
-
Ah OK
Like this only I used 2sk208 gr
Posted 1st December 2016 at 01:20 AM by nige2000
-
Don't those FETs have rather tricky availability? Do a similar thing but using bipolars, its easier and cheaper. But it makes no sense to put the load across the current source, the load should be across the active transistor. Putting the load across the current source makes the circuit ps sensitive.
If you use bipolar transistors then you can make a buffer which can handle 1VRMS with two LiIon batteries for supply (7.2V).
Notice the JFET is needed to get zero input offset voltage because its fed by a pot (pots are very DC sensitive), but with no pot the input offset isn't very important. The FET needs an output cap all the same as its Vgs is not going to be zero.Posted 1st December 2016 at 01:24 AM by abraxalito
Updated 1st December 2016 at 01:36 AM by abraxalito -
There's loads of 2sk208 FETs about though
Been ps sensitive isn't much of an issue for me in this instance as the power supply is already there in my amp
I think I'll have to try both
I think the pot needs to be after buffer not before as it is now
If the pot is DC sensitive it would either need a traffo or DC blocking cap before signal enters a pot?
I'll definetly try a tvc on amp out too as I think it would be nice not to use a potPosted 1st December 2016 at 01:56 AM by nige2000
-
You mean you'd use the amp's power supply to feed a buffer? 32V either side? Or you have another one? If you tap off from your batteries that'll have a fair bit of load-dependent noise on it, wouldn't want to feed that into a line stage.
Yeah you either want a trafo or a cap to ensure no DC on your pot, preferably replace it with TVC.Posted 1st December 2016 at 02:12 AM by abraxalito
-
yes tapped of the batteries at +/-19.8v
i tried other supplies for the buffer for comparison at the time, and prefered the taps of the cells by quite a margin, i had assumed they were able to mostly absorb whatever noise the amp made
i think the only way to figure it out is to build an amp/dac your way for comparisonPosted 1st December 2016 at 05:20 PM by nige2000
-
Those cells have ESRs in the region of a few milliohms. So with six of them in series you'll be into say 20milliohm. Then with output currents peaking at say 2A that's 40millivolts peak of signal-correlated noise on your rails. Whereas for regulators they put out in the region of tens to a hundred uV of noise.Posted 1st December 2016 at 11:37 PM by abraxalito
-
i have two amps the lifepo4 one and its predecessor which is very similar a dual mono version audiosector with a shunt supplied buffer dcb1
actually was the point of the experiment to build a battery version of it for a more apples to apples comparison
but there is no comparison the lifepo4 version is miles ahead
sure you could pick flaws all day long with both
but still remains a reasonable comparison for ps i thinkPosted 2nd December 2016 at 12:53 AM by nige2000
-
Yeah, passive ps beats active any day in my experience.Posted 2nd December 2016 at 01:04 AM by abraxalito
-
moved the pot on the amp from pre buffer to post buffer think its substantially better, cleaned up the top end a bit too
was playing with line level trafo 15k/15k, seem to have more soundstage or atmosphere
also thought there was maybe a slight smoothing effect, on the ab testing
possibly why it works well with sigma delta dac chips?
get the feeling a slight step down may negate any possible negative effectsPosted 7th December 2016 at 12:49 AM by nige2000
