Pulse transformers SPDIF/AES-EBU

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,

We can read at the beginning of the post than a cap would be better sometimes for sound quality than a pulse transformer for the DC stopping function.

What is the best according to you if the length is less than 20 cm between the spidf source and receiver with a BNC connectors and 75 ohms cable?

1) - Stopping the ground loop with a pulse transformer. Is there a minimum 1:1 pulse transformer quality to avoid the problems seen in the first posts ?

2) - Allow the ground loop with sharing the Gnd of the two pcbs (when it's possible) and only use a cap for security as DC blocker, because we can hear it (more details or bass resolution - see the first posts- ) against pulse transformer ?

If 2), I read than the maximum pulse fhz of a spidf signal is 6M hz : what sort of cap would you use ? The just enough band pass width you need for it or the fastest and stable like COG-NPO or polystyren?

A) smd or bulk ?
B) bulk X serie (military) COG ceramic 1uF - 5 mm between legs, plastic case
C) Polystyren bulk like 100 nf MIT for speaker crossover
D) Film : MKT,PEN, MKC, MKP
E) bypassing using A) to D)

Sorry if two cents, thanks
 
adrichia said:
I do understand the reflection when impedance isn't properly match. But I appreciate u can answer each of my question. That will be great.
The coaxial cable n bnc plug is 75 impedance. Now by shorting 350 ohm instead of 75 ohm. Is it acceptable?
No. The fact that you have asked this question shows that you do not understand reflection.

You need to do a lot more reading and experimenting before you will be ready to 'improve' a digital output interface. You will also need some serious test equipment and the knowledge to correctly use it.

I would be cautious about 'improving' digital circuits, as I only have 45 years experience of RF and audio circuits and only have a fairly standard 20MHz analogue oscilloscope.
 
Hi Eldam, thanks a lot for some input. Giving more questions instead :)
The best approach that I am using now is the improvise spdif I circled in the picture. I am relying on the Dac end cap to do the blocking.

Yes the ground is carry over through the coaxial shield to the dac, which has a proper ground. Strangely, the improvise spdif will not work if shield is not grounded to the CD player.

If u could answer my questions on my initial post, it will be great. Rest assure, I read the beginning of this thread couple of time already. Like I said, my knowledge is lacking, so I need u guys to help me out.

The transformerless spdif is meant to take a better square out of standard digital out the player can offer. There has been graph in the web showing how bad the transformer is doing to the signal.
 
hi Adrichia,

I'm just a noob, sorry I can't help. My question is near yours because I will use some old dac chips I have with a spidf board using the Wolfson receiver ship. As it will work with two pcb I suppose myself than the question of ground loop is quite important.

But I'm not planning to improve anything but just to understand where i have to go both the money and my understanging as it seems there are two solutions : with and without pulse transformers. Some very good DAC dont use pulse transformer. I go out with a simple coax spidf from my source

The few things I believe to understand are (but don't know really) :

- a very good pulse transformer is better than without, but are expensive if you want a good enough quality to avoid the problems seen in the first posts here.
- I'm not sure that they are able to stop the DC and at the end a DC bloking cap have to be use. In this case the benefice is just to have not a big ground loop as the GND is stopped by the pulse transformer.

- as the spidf fhz wave is between 100 K hz and 6 M hz (or 19 Mhz with 24 bits/192 hz ?) and the impedance line is 75 ohms : the DC blocking cap have to be > at 100 nf (because a serie cap is a high pass filter and 100 nf give 106 K hz for the lower fhz !). So at least 220 nf seems to be better but I really don't know !

- parasitic inductance seems to be important for a cap in serie for signal coupling and in this case a smd cap should be better than a bulk one... at the end the choice of dielectric for loss is important and the polystyren in theory is the best for that !

- piezzo electric effect (noise with stress and micro moves of the caps pn its pcb) of smd can be a problem and maybe bulk capacitor is better ?

It is just the state of my understanding but not the truth, not sure it can help you ! Sorry if you think I high jack your question, I am like you, lost between question and practical applications. I think DF96 is right : firsdt you have to try to understand what you do before going to practice. I am at this first point.

regards....
 
adrichia said:
There has been graph in the web showing how bad the transformer is doing to the signal.
A transformer does not go down to DC, so the waveform looks different. Fortunately, this does not matter at all - it has no effect on sound quality. Transformers have advantages, such as galvanic isolation, which is why they are used. As a general rule, nobody adds wound components to a circuit unless they are necessary so if they are present you can be reasonably sure they are doing something useful.

Note that this is for commercial equipment; some DIYers sprinkle chokes and transformers randomly.
 
Hi Eldam,
Thanks for sharing. My main concern here is whether I've short too much power to the ground. If Leg 2 is outputing 5volt, and voltage divider is 550 in series and 350 to ground.. mean 1.9volt in series, while 3 volt short to ground? May sound silly, but appreciate you can help out.

Hi DF96,
Without a doubt, the original circuit serve a very important purpose as you have explained. Just like a combustion engine car, where a very fine restrictive air filter is a must because the car is meant to operate in various kind of imaginable condition; that's the design principle every car manufacturer must adhere.

Now just consider, the environment has been best controlled, and the variable has been made singular, only 1.

What is the best method moving forward? Am I shorting too much voltage to ground?

Thanks for being patience with me and I am looking forward on your knowledgable inputs.
 
Hi,

i make a mistake : CD723 use SAA7378GP before the dac chip

But : Philips CD710/721/722/740/750/751 use the SAA7345GP, some of those use the pin 2 as EBU output and resistors for adaptation impedance and most of time a little shelded trafo who can seting up by screw on the top of it. HP filter or LP filter (I don't know) are used maybe with a 1 uf caps if my memory is good...

Maybe you will find some shematic on lampizator site : it talks about modding the spidf output of some Philips & Sony players : check it.

this one can help you like the vasily list : The complete d/a DAC converter list - DutchAudioClassics.nl

good luck,
 
Last edited:
Hi,

I design RF and high speed data lines for a living and I would never put a transformer on such a line if it wasn't absolutely necessary for isolation reasons. Careful PCB layout, a good line driver, NP0/C0G coupling caps is a good place to start. If it is only for diy I would also skip the low level of spdif and keep the signal at whatever level it is and then take it down to whatever level is required by the receiver right at its input. If not using BNC I Woulrd recommend a cheap RCA rather then an expensive one, the reason being that the cheap ones are shorter and hence, presents a very short impedance discontinuity that will have a very limited influence on the transmitted signal.

BR,
Anders
 
..NP0/C0G coupling caps is a good place to start.

Thank you for the good hints.

Does it need more than 100 nF for coupling (DC stopper)?

My understanding is the EBU's frequency wave beginn around to 100 k hz to stop near 6 M hz. So with a 75 ohms line impedance, A little more than 100 nf as coupling cap is enough (let say 200 nF...). Does it need more : I have 5 mm spacing leads bulk 1 uF COG but would you advice only a smd Cog of 100 nF or the both in // ?

If it is only for diy I would also skip the low level of spdif and keep the signal at whatever level it is and then take it down to whatever level is required by the receiver right at its input. If not using BNC I Woulrd recommend a cheap RCA rather then an expensive one, the reason being that the cheap ones are shorter and hence, presents a very short impedance discontinuity that will have a very limited influence on the transmitted signal.

that's what I saw on a chineese DAC between my Squeeze Box and CS8012 spidf receiver : the signal & gnd was directly put on the receiver (no change on the side of the squeeze box : its spidf has a cap in serie before the rca) : result was better : see it on the Lampizator site.

We often read than the line with spidf must be at least 2 meters for avoid reflexions, is it true or did you advice the shortest distance when spidf cable is used (when DIY with one of the devices or DIY the both (emitter/receiver)?

Do the 75 ohms plugs make a real difference against light RCA plugs in both scenario: with real 75 ohms line and with DIY no attenuation line (OP question) in the case of the receiver is strong enough for the voltage ? (= do the ears are able to listen to a difference?)

thank you for your help, a lot of questions, I'm sorry but I can always hear the differences when i skip a spidf cable with an other one... want to minimize this "problem". Advices are welcome :).
 
Last edited:
We often read than the line with spidf must be at least 2 meters for avoid reflexions.....:).

that seems like good idea, unless you have a transmission line that is terminated correctly at both ends. So, in general, keep the length to > 2m.

RG6 quad-shield cable works really good, so my suggestion is to at least try this cable as a cheap and excellent solution.

Nick
 
Hi Eldam,
Yah, that is where it got me started on this transformerless SPDIF mod 1 year ago all because of the generous sharing at Lampizator site.
I have concern only on the amount of volt/current shorting back to the player, with the requirement to maintain the existing original circuit of the player's digital out.
The not so good part about SPDIF is that, it is real time with "non-genuine" data correction method, i believe some algorithm re-constructed data sound better than the actual data and vice verse. As such, it is best we make the spdif output as "clear" as possible to reduce interpolation filter from playing the algorithm trick~


Hi Bappe,
Thanks for the re-assurance. I read something similar that those transformer and isolation is not really that critical when the transmission is of minimum distance. Like our normal home use. Where everything is not more 2 meter apart. EMI, RFI are to only intervene without that short distance. The idea is to do it correctly. Appreciate you help answer the questuions I posted in page 2. Thanks.
 
Thank you for the good hints.

Does it need more than 100 nF for coupling (DC stopper)?

No, 100nF is quite enogh.

would you advice only a smd Cog of 100 nF or the both in // ?

One is always better than 2, the key here is to keep it physically small for low inductance.


that's what I saw on a chineese DAC between my Squeeze Box and CS8012 spidf receiver : the signal & gnd was directly put on the receiver (no change on the side of the squeeze box : its spidf has a cap in serie before the rca) : result was better : see it on the Lampizator site.

Lampizator is doing some really strange things and i would PERSONALLY stay away from most of it when it comes to data transmission.

We often read than the line with spidf must be at least 2 meters for avoid reflexions, is it true or did you advice the shortest distance when spidf cable is used (when DIY with one of the devices or DIY the both (emitter/receiver)?

From a signal integrity pov for the transmission, there is no need for a longer cable. If the transmission is generating reflexions (it allways does to some extent) it might be a good thing if the first reflexion does not reach the receiver when the gate is still open. A common "gate open" time is around 8nS and this would practically point towards a cable that is at least 1m or so. This is of course valid only if the main reflection is from the ends of the transmission line, if the main sources of reflexions are from both tranmitter, receiver and connectors or even the cable itself then its all just a lottery :)

Do the 75 ohms plugs make a real difference against light RCA plugs in both scenario: with real 75 ohms line and with DIY no attenuation line (OP question) in the case of the receiver is strong enough for the voltage ? (= do the ears are able to listen to a difference?)

Lets say that we need a bitrate of max 12Mbit (32*2/192k), this translates to a BW of 6Mhz, or does it really? The figure 6MHz is frequently said to be the analog BW needed for spdif but it is not. This is only the fundamental and we need around 5 times this to be able to get good signal transmission. This is sort of a rule of thumb for digital transmission but we do not want to just get by, we wan very good transmission with very little jitter contribution, right? If we transmit this 6MHz signal on a 60MHz BW line we can get the 9th harmonic in and this will give us a very nice and sharp rise and fall of our signal, going beyond this is just a waste of time and dollars :) It is the rise time of the signal that dictates the needed BW, not the fundamental transmission rate. So, now we know that we need an analog BW of 60Mhz in our line, a 6MHz square wave on a 60MHz line gives a rise time (10-90%) of app 6nS. In general one can get away with around 25mm of impedance discontinuity for every 1nS of rise time in digital systems, in our case 150mm. This is not valid for SPDIF since it is allot more sensitive to timing issues ie jitter. The WORST CASE jitter contribution for a discontinuity alone is related to twice its length or about 10ps for every 1mm. So use a 75Ohm connector, and if not keep it short!



thank you for your help, a lot of questions, I'm sorry but I can always hear the differences when i skip a spidf cable with an other one... want to minimize this "problem". Advices are welcome :)

No sweat :) diy = sharing

edit :
SPDIF is a crappy interface and the only way to get rid of its problems is to have a separate, independent clock for the DAC. No wonder that cables make a difference as it is today with HW that is not up to standard there is no telling which cable that will work best :(

BR,
Anders
 
Last edited:
Thank you Anders for this serious explanation : all what I woulk like to know for a long time : I know it now :) and have a road map.

Thank you Extreme_booky, I have already 3 x 75 ohms digital Audio branded cables :eek: (shamed...bought in the 90s') : I will try simple & cheap :) TV 75 ohms quaded shelded to have experiment with short and long length with or without connector in the way Anders explain just for didactic experiments.

Thanks for sharing gentlemen :cheers:


P.S. : Adrichia : Sorry for this little high jack. Did you check the max voltage input of your receivier chip to try your spidf line with your genuine source configuration without shunt to gnd ? Is a simple wirewound Z foil Rhopoint resistor as voltage diviser a bad idea ? Too much Johnson noise ?
 
Hi bappe, thanks for the great knowledge sharing.

Hi Eldam, nothing to hijack here, it is not started with me anyway :) the dir9001 receiver can take the full voltage from saa7345. Just not feeling comfortable letting it transmit without any protection. Not sure what kind of weird electrical effect if to run with just a resistor in series.

It is always great to see ppl doing feed backs.
 
Disabled Account
Joined 2002
The circuit seems OK. I think it is good that they RF decoupled the output GND. How is the quality of the most important part (transformer) and how does the waveform look on an oscilloscope ? If you remove that one you won't have galvanic isolation. I know many web sources claim that is not necessary but I have experienced quite a few times that galvanic isolation is making things better. This is especially valid with a source like a Squeezebox Touch. I can tell our Subbu V3 needs a transformer when coupled to a SBT. Same counts for other DACs I tried with a SBT.

Upping the 100 nF tot 220 or 470 nF would not hurt.
 
Last edited:
A Little Sussumu thin multilayers resistor....;) which those values near the input receiver for the reflexions ? Is it the goal here ?

With 9 harmonics, if my calcul is good we have an electrical bandwidth between 100 k hz & 4 G hz... I'm asking what is the effect on a resistor with such signals? No bad with good low value as I see one between the 50 000 hz crystal and the DAC on the subbu (i suppose a ferrite bead of the same ohmic value has an another effect... ?!)

As a transformer stopping the dc on the gnd also to allow two ground loop (each on its own device): is it possible to put the equal resistor on the Gnd wire as well ?

Another subject : i remenber with the SAA75 or SA78xxx about the I2S when you remove all the RC and the caps nears the dac chip : the sound is cleaner with more resolution... maybe because of the bad quality of components on the entry level CD player 15 years ago (like the Philips cd723 which was a good player for the price and were improved with betters parts and some tweaks).

Is a good measure with scope on this case (spidf) can give a bad result with ears... and vis verca ?
 
Last edited:
jean-paul,

Why a 270, 470 or more value for caps as DC stopper can not do a prejudice here. I believe we must read here than such values are better than the 100 nf which theoricaly and experimentaly (scope) is enough ?

Would you say than the low end of the filter is important ? : bandwidth < at 100 000 K htz with spidf ? Are you suggesting than for ears it's better as Gary advise for example ... for ears I mean ?

My understanding :

- if I take 75 ohms : 102 nf is enough as it is a high pass filter, all under 100 K hz is removed with 6DB octave attenuation.

- if I take the ohmic entry impedance of the chip receiver (like Gary with its 470 k ohms with the W8804 chip receiver), the calcul is not he same ?

- Is there a third way (that serious people calcul with scope with each design) as for example take the RC of the line where R is the shunt resistance with each pcb and line to the Gnd ?)

I asked above if a resistor can help for reflexion (short line or line non 75 ohms) for signal &/or gnd wire. In the same spirit : is a cap needed on the Gnd wire ? Is there any garbadge sound on these under 100 K hz or none as it is spidf bandwidth signal ?

Sorry for the naive questions ? c'est juste passionnant !
 
Hi bappe,

I read some papers from Murrata.

I understand than the lower case is better for inductance as with the nine harmonics our spidf signal klimb to 4 Ghz (with 6 Mhz as fundamental).

You talked about harmonics. Is there sub harmonics too (below the fundumental) ?

COG/NPO are better for HF. But all the little smd caps with years can make some cracks with solder. there are also little piezzo electric effect and micro movements on the size of the caps and flex of the pcb which seem to make noise. I mean here hearable noise (20 hz-20 khz) ?

Is it important for EBU signal in audio device or not ?

Anyway, I understood than the low profile of the case if much more important (inductance).

My question is in fact about some special smd caps from Murrata : MK & J8 series which are "antinoise"... not in the way of hearing but in the way of electrical solutions (before a processor in computers). Does these special caps have an audible effect in serie with a signal like spidf ?

But I can not find it on mouser or Digikey !

Pfff all those questions that normal people like me have since Prometheus gave the fire to the human beings ! Et voila, we have DACs & atomics centrals :D ; BTW, audio is a terrible Pandora box... (what ! 2 cents of culture is not bad :drink:)
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.