Hi all: Tired of searching entry level info about conditioning PWM DAC signal using a tranfo instead a low pass filter :-(... Now I need some help pls.
Taking as base a Marantz CD6000 (DAC stage similar to long tweaked Marantzes CD6x), I'm trying to replace the entire output stage (based on HDAMs or HDAM+Active LPF depending on CD6000 model).
My aim is using a line signal trasnformer to both low-pass filtering PWM DAC output and converting from differential to single ended output, but I have some doubts (because saddly I'm not electronics engineer), so any light over them will be really appreciated:
I sent an email to Lundahl who fast and gently answered me suggesting the usage of their LL1527 trafo, however its bandwidth is 10Hz-150kHz. Searching in their catalog I find 5402 being more suitable to my needs: in fact its bandwidth is 20Hz-40kHz. Am I radically wrong when thinking that my target bandwidth is 20Hz-20kHz? Why they offered me a model with a higher bandwidth?
As you can see I'm really confused, so any help, any usage example, any addressing to info will be really really appreciated.
Taking as base a Marantz CD6000 (DAC stage similar to long tweaked Marantzes CD6x), I'm trying to replace the entire output stage (based on HDAMs or HDAM+Active LPF depending on CD6000 model).
My aim is using a line signal trasnformer to both low-pass filtering PWM DAC output and converting from differential to single ended output, but I have some doubts (because saddly I'm not electronics engineer), so any light over them will be really appreciated:
- Taking the output from a NPC SM5872BS Sigma/Delta DAC (16 times oversampled 44,1kHz digital signal), should this signal be dropped directly to the transformer or, as I think, might it be previously low pass filtered, say, with a RC filter?
- In both previous named low pass filter and transfo, the target bandwidth should be as closest as possible to 20Hz-20kHz? I assume that it is true.
I sent an email to Lundahl who fast and gently answered me suggesting the usage of their LL1527 trafo, however its bandwidth is 10Hz-150kHz. Searching in their catalog I find 5402 being more suitable to my needs: in fact its bandwidth is 20Hz-40kHz. Am I radically wrong when thinking that my target bandwidth is 20Hz-20kHz? Why they offered me a model with a higher bandwidth?
As you can see I'm really confused, so any help, any usage example, any addressing to info will be really really appreciated.
You want as wide a bandwidth as possible and a secondary RC filter that operates way above the audio band. The LL1527X might be fine but I dont know anything about your chip except the datasheet info. It would seem that the chip will only supply 1ma of output current @ about 4.7V so 5kohms would be the minimum value of the load impedance. A test CD and scope would be the best test.
In general, trafos work great with V out chips as long as you keep the cables short.
In general, trafos work great with V out chips as long as you keep the cables short.
Thanks for your repply Bill.
Based on what you point both trafos should work, however it's true that 20Hz base frequency on LL5402 might produce an attenuation on lower frequencies and that LL1527X should work better if secondary RC filter deals with 20kHz-100kHz bandwidth filtering.
In any case I think I understand your concept: transfo deals with a first filtering stage and differential to single ended conversion, then RC filter deals with a second filtering stage, it's it?
Based on what you point both trafos should work, however it's true that 20Hz base frequency on LL5402 might produce an attenuation on lower frequencies and that LL1527X should work better if secondary RC filter deals with 20kHz-100kHz bandwidth filtering.
In any case I think I understand your concept: transfo deals with a first filtering stage and differential to single ended conversion, then RC filter deals with a second filtering stage, it's it?
I'm pretty sure the NPC DACs don't use analog filtering (i.e. switched capacitor) so without a bunch of filtering you'll get a big chunk of noise coming through. Marantz IIRC use 4th order filters (I may be wrong) so a transformer on it's own, even one with a low bandwidth, may not be enough. I'd hunt down a circuit that provides 4th order filtering at around 40-60KHz. Look through transformer manufacturer's datasheets for something that might do.
True, this DAC outputs signal as PWM, so filtering must be implemented outside the DAC. That's why I love this kind of DACs, because of the flexibility of doing that filtering as you want
In fact CD6000 uses a 2nd order RC low pass filter followed by an active one. DAC datasheed shows two practical applications: one based on a passive+active and other based on an active+active low pass filter.
So your advice seems to be in line with Bill's one: transfo + additional filter. In fact Heart's audio tube based CD6000 mod uses the onboard passive filter and then a transfo prior to amplify the signal. My doubt now is knowing experiences in this line, ie, used transfos, schematics, etc, for similar applications.
Thanks for your help, Spartacus.
Checked the datasheet and yep, NPC use a second order filter. Doesn't say what the -3db is though. I guess you could try it with and without extra RC filtering and make a decision by ear. For a transformer I'd use one with a bandwidth as low as possible without significantly impacting the audio range.
Oh God, this is really a headache for a electronics absolute fool like me 
Let's state that -as Bill pointed based on DAC specs- DAC output drops 1mA @ 4,7V, what forces a matching impedance > 5kOhm in the transfo primary.
After reviewing Lundahl catalog I find two models with freq responses fit to my needs:Lundahl LL1540 (input line transfo) and LL5402 (output line transfo)...
... and here I'm lost with specs: no-load impedances, static resistances...
I'm not able to understand if one of these models fits to my needs. I just assume that matching impedance must be achieved adding a resistor between primary and - (or so I think).
Again, any help about DAC-tranfo integration will be really appreciated.
Let's state that -as Bill pointed based on DAC specs- DAC output drops 1mA @ 4,7V, what forces a matching impedance > 5kOhm in the transfo primary.
After reviewing Lundahl catalog I find two models with freq responses fit to my needs:Lundahl LL1540 (input line transfo) and LL5402 (output line transfo)...
... and here I'm lost with specs: no-load impedances, static resistances...
I'm not able to understand if one of these models fits to my needs. I just assume that matching impedance must be achieved adding a resistor between primary and - (or so I think).
Again, any help about DAC-tranfo integration will be really appreciated.
Yeah, it gets confusing. To simplify it somewhat, don't use a trafo wound for input duty. The construction is vastly different and complicates the implimentation greatly. Some will argue with this but it is true non the less.
Output trafos are designed to operate without a specified load so the impedance of your preamp is reflected into the primary. It will be on the order of 50k-100kohms, plenty high for the chip. An output trafo's rated impedance is just that, the load it CAN operate into and meet specs, not a required load.
Whether you use a limited bandwidth product or not is up to you, but I would not. I would merely use a simple 1st order filter before and after the trafo
Output trafos are designed to operate without a specified load so the impedance of your preamp is reflected into the primary. It will be on the order of 50k-100kohms, plenty high for the chip. An output trafo's rated impedance is just that, the load it CAN operate into and meet specs, not a required load.
Whether you use a limited bandwidth product or not is up to you, but I would not. I would merely use a simple 1st order filter before and after the trafo
Hi Bill: I've been looking for that trafo at Jensen but it seems not being in their catalog. Anyway I've found an application note from Jensen very similiar to what I wanna do.
It seems using a low pass filter, as you recommend, prior to goint for the primary. Just one problem: JT-11-SSP costs 140$ (glub)
I'll try, again, the Lundahl approach and I'll follow his and your advice: using the ll1527 and not filtering in the transfo.
Bill, your help is really appreciated. Thanks
It seems using a low pass filter, as you recommend, prior to goint for the primary. Just one problem: JT-11-SSP costs 140$ (glub)
I'll try, again, the Lundahl approach and I'll follow his and your advice: using the ll1527 and not filtering in the transfo.
Bill, your help is really appreciated. Thanks
It seems a good polivalent transfo, very suitable for experimenting and using in different configurations... claims to try mantaining input impedance as low as possible, for this application it might be seven times higher than the nominal 680 Ohms, so the result is not predecible unless tested.
Link to Jensen JT 11-DMxx datasheet
Link to Jensen JT 11-DMxx datasheet
You might want to download the Jensen article by Bill Whitlock about audio transformers and read it a few times. It will help you wade through the choices and give you a better perspective.
The dac chip will never see a 680ohm load until you get below the threshhold where inductance affects the low freq.cutoff more than the reflected impedance does.
The chip probably has a source impedance around 500 ohms, going by the general rule of the 10/1 ratio of rated load to source. The trafo will just reflect the secondary load until you approach the area where the trafo inductance dominates the perceived chip load, and that will be in the subsonic range below 20hz. When the source impedance equals the perceived load impedance you have a 3db voltage loss, and as the load impedance decreases at lower frequencies the losses become much greater. That is why you want a lot of inductance, the more you have, the lower the low freq cutoff will be.
As long as you are transferring voltage, not power, the lower the source, the better. Power transfer involves impedance matching, but that's not what we are doing here.
The dac chip will never see a 680ohm load until you get below the threshhold where inductance affects the low freq.cutoff more than the reflected impedance does.
The chip probably has a source impedance around 500 ohms, going by the general rule of the 10/1 ratio of rated load to source. The trafo will just reflect the secondary load until you approach the area where the trafo inductance dominates the perceived chip load, and that will be in the subsonic range below 20hz. When the source impedance equals the perceived load impedance you have a 3db voltage loss, and as the load impedance decreases at lower frequencies the losses become much greater. That is why you want a lot of inductance, the more you have, the lower the low freq cutoff will be.
As long as you are transferring voltage, not power, the lower the source, the better. Power transfer involves impedance matching, but that's not what we are doing here.
Bill, you're THE master
Some time ago I read about speaker impedances and why they might match with amplifier. Few days ago I read about line level and they told that line input and output impedances might be far from being similar. It drove me confused because it was absolutely opposite to what said about speakers and amplifiers.
Thanks to your explanation I know that when dealing with speakers (power) impedances must match and why they don't do it when transfering voltage.
It could seem a fool thing for an electronics engineer, but I am only a computer scientist trying to self-learn electronics, this kind of help is really very appreciated.
Thanks Ed for participating on this thread. I hope we could collaborate since it seems both you and me are pursuing the same -or very similar- target. Just one question Ed: different DAC but sigma/delta output, or different DAC with voltage/current output?
I'm sure the ideas from this thread will be very useful for the tweaking community. Thanks all.
I add the link to the article about transfos pointed by Bill: Audio Transformers by Bill Whitlock
I'm sure the ideas from this thread will be very useful for the tweaking community. Thanks all.
I add the link to the article about transfos pointed by Bill: Audio Transformers by Bill Whitlock
Last edited:
Bill, you're THE master
Some time ago I read about speaker impedances and why they might match with amplifier. Few days ago I read about line level and they told that line input and output impedances might be far from being similar. It drove me confused because it was absolutely opposite to what said about speakers and amplifiers.
Thanks to your explanation I know that when dealing with speakers (power) impedances must match and why they don't do it when transfering voltage.
It could seem a fool thing for an electronics engineer, but I am only a computer scientist trying to self-learn electronics, this kind of help is really very appreciated.
I'm sure no master, by any stretch. A guy like Bud Purvine IS a master. I'm not exactly technically accurate in my ramblings, just trying to be helpful.
...and it was as if I had run into the door jamb.
I think the latter...You illuminate another area I don't know enough about. When I am confronted by my own lack of knowledge I retreat to (or go find) the references to help me along the way.
I try to improve the singular elements of my system. In this case the cd player/dac. I have the shigaclone, I need the dac.
Last edited:
Ok, you mean DAC as a device while I mean DAC as a substage on a CD player. Have you considered any approach? I'm not very experienced in that matter but I remember Lukas Ficus (from Lampizator) combining a Wolfson test board with a tubed output and saying it was the best DAC he had heard is his life.
The link: WolfsonWM8742DAC
Anyway I'm interested in building a decent DAC (as device, not as player stage) since I've decided moving, after finishing this project, from CD players / transports to a WiFi or memory card based transport.
The link: WolfsonWM8742DAC
Anyway I'm interested in building a decent DAC (as device, not as player stage) since I've decided moving, after finishing this project, from CD players / transports to a WiFi or memory card based transport.
You might find this interesting. Decware is now selling their version of the Ebay Chinese upsampling CS4398 dac board installed in a cheap case, for $875. I have had 2 of these boards for quite some time using output trafos with both. They cost $90 on Ebay and they are quite well built and sound excellent with trafos.
Decware has a lot of nerve, but proves a lot of the Chinese stuff is quite good. I have one built into a Redbook player and feed the other with an SB3.
Decware has a lot of nerve, but proves a lot of the Chinese stuff is quite good. I have one built into a Redbook player and feed the other with an SB3.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.