Signalyst DSC1

I received a bunch of 74AHCT595 chips made by Diodes Inc. instead of by TI. Same problem. I am doing only careful hand soldering and have had more fail than survive. One that survived the soldering had a channel fail later.

The 74AHCT595 chips I'm using are by NXP.

How do you do the soldering? I'm using RoHS compliant SMD soldering paste and hollow head / cup tip soldering iron, and IIRC the Weller solder station is set to 350C. Then I just pull the iron over each pin and clean solder flux with cleaning spray. Very small amount of solder paste is enough, otherwise it easily gets under the chips/behind the pins. And the paste must be at the outer tips of the pins to avoid residuals left in any improper place.
 
For sure SN74AHCT595 from TI, available also in PDIP package, appears more realiable.
But I don't know if operating frequency (max 110 MHz) or other features are appropriate

Where do you source PDIP variants? I would have used PDIP instead of SOIC if I would have been able to ensure supply of needed chips, because soldering PDIP is much easier...
 
74AHCT595 Chips Still Failing

My guess is that all 74AHCT595 SOIC chips come from the same Chinese factory no matter the company. I suspect a bad batch coming through both Diodes Inc and TI. All 16 of the chips that came with the Chinese kit survived my soldering. I am only about 6 out of 20 with chips that I ordered recently. I replaced two bad 74AHCT595 chips two more times and had some output channels fail each time. The NXP 74AHCT595 is not available in SOIC.

I redesigned my PCB's to provide only 5V input signals so that I can use 74VHC595 chips which I was successful with in the past. These PCB's will also accept the 74AHCT595 chips. I will try both in my recalibrated reflo oven in a couple of weeks when the PCB's arrive.
 
74AHCT595N DIP availability

Signalyst,

I have been using the Dip version of the Texas Instrument 74AHCT595N in my version of your DAC without any problems. I got them from Digi-Key,
P/N 296-33695-5-ND.

See Page 13, post 128 of this forum above for a picture of the beast.

N.
 
My guess is that all 74AHCT595 SOIC chips come from the same Chinese factory no matter the company. I suspect a bad batch coming through both Diodes Inc and TI. All 16 of the chips that came with the Chinese kit survived my soldering. I am only about 6 out of 20 with chips that I ordered recently. I replaced two bad 74AHCT595 chips two more times and had some output channels fail each time. The NXP 74AHCT595 is not available in SOIC.

I redesigned my PCB's to provide only 5V input signals so that I can use 74VHC595 chips which I was successful with in the past. These PCB's will also accept the 74AHCT595 chips. I will try both in my recalibrated reflo oven in a couple of weeks when the PCB's arrive.

Strange, here's the DigiKey packaging label for my NXP 74AHCT595 (from Thailand) batch:
shift-reg.jpg


I'm using soldering iron and not oven, so the chips don't really get hot. But of course shouldn't make difference, reflow is supposed to be OK.
 

Attachments

  • DSCN3461.jpg
    DSCN3461.jpg
    1 MB · Views: 765
16-bit Sine-wave DSD Working!

See pictures of my latest differential 16-bit sine wave weighted DSD and Amanero Interface-Mute boards.

First listening impressions are promising. I first connected the outputs to my Cinemag transformers. Very nice sound, but lacking a bit in bass and musical energy. No noise that I can hear. Much more detail than the Chinese 32-bit version.

Next I sent the output to a JFET buffer board and then through the Cinemags. Bass is powerful and deep - best I have ever heard. I use this same JFET buffer circuit on the outputs of my reference DDDAC NOS PCM1794 DAC which gave a similar sonic improvement. Lots of musical energy and detail resolution. Also, lots of new parts sound which is very bright and edgy. In a few days of continuous play I will know how good this DSD NO-DAC can sound.
 

Attachments

  • DSD Diff-2.JPG
    DSD Diff-2.JPG
    958.2 KB · Views: 747
  • DSD Diff-1.JPG
    DSD Diff-1.JPG
    800.2 KB · Views: 715
Good to see your progress Carlsor, it looks very promising.

I've been short of time to progress my projects after I got frustrated at damaging my Chinese diff board and I've been waiting to see how you're project progressed.

Assuming it all works out, are you willing to share your boards/schematics?

What DSD rate are you able to run your Amanero at? As I have had some success at running my Buffalo 3SE DAC at DSD512 via a Beaglebone/Botic/NAA I will be looking to develop that input interface.

Cheers

Ray
 
Some 16-bit DSD project Details

I will share schematics, PCB's, or Eagle files if anyone wants to attempt this rather complex project. See picture of blank PCB's. This is NOT a beginner project.

I committed the same sin of using tiny IC's for U3, U8, and U9 as the Chinese DSC1 because no larger size was available in the 74AHCT logic family which can accept 3.3V family inputs and output 5V signals - especially for the XOR "86" logic gates. I used a solder paste stencil and reflow oven to install all the SMD parts on the 16 Bit PCB. I hand soldered the Amanero interface board. Not a single failure of the outputs of the eight 74VHC595 IC's.

I am using JRiver 22 at 2XDSD with 44k/16 source material. Higher rates are not possible in my current setup. Theoretically, putting a 6mhz DSD through a 16-bit sine wave "digital filter" should act like a 375Khz LPF with constant phase shifting. Higher DSD rates would push the effective LPF much higher and may do more harm than good.

The only HF roll-off filtering after the 16-bit sine digital averaging is a 220pf WIMA FKP2 across the outputs, the Cinemag transformers, and a 150pf WIMA FKP at the output of the Cinemags. I also have a highly modded Behringer DCX 96/24 DSP crossover-equalizer before my speakers. I biamp using the low and mid outputs which go up to 20Khz. All musical content above 20Khz goes to the high section. My DDDAC NOS PCM1794 uses the same buffer and Cinemags as this DSD, but the high channel consistently shows some musical content above 20Khz. This DSD has yet to register any music above 20Khz on the Behringer display which implies that this DSD design is doing a considerable amount of HF filtering.

There have been many DIY DSD design projects that start out promising and then fade away when sq improvement stops short of musical nervana. I am trying to temper my enthusiasm here until the break-in is complete. Hopefully there is no residual digital hardness that is being masked by the break-in edginess.
 

Attachments

  • DSD Diff-3.jpg
    DSD Diff-3.jpg
    382.5 KB · Views: 1,041
Audio Systems

Bass is powerful and deep - best I have ever heard
-----which of course applies only to my stereo systems which are as follows:

Preamp is an upgraded "Truth" buffered LDR volume attenuator and source selector. Outputs go to both the Headphone and floor-standing speaker systems.

Headphone system consists of a highly tweaked DIY AMB M3 headphone amp and Grado GS1000 Headphones.

Speaker system consists of biamped Emerald Physics CS 2.3 OB speakers with a highly modded Behringer 2496 DSP crossover-equalizer, Tortuga LDR volume control with DIY JFET Buffers after the Behringer, DIY 10w/ch Moskido tube-MOSFET hybrid amp for the mid-highs, and a 200w/ch Marsh A400S for the lows.
 
See pictures of my latest differential 16-bit sine wave weighted DSD and Amanero Interface-Mute boards.

First listening impressions are promising. ...
... Lots of musical energy and detail resolution. Also, lots of new parts sound which is very bright and edgy. In a few days of continuous play I will know how good this DSD NO-DAC can sound.

Good news, Carlsor !

I will look forward to hearing from you.
 
16-Bit Sine Differential DSD NO-DAC Update

Good news and bad news. First the bad news - my inner ears are full of sinus fluid so my hearing is a bit impaired at this time.

The good news is that this DSD no-DAC is very listenable. I can listen for hours with no fatigue. Bass is still deep and tight. No noise. Lots of musical energy. I can hear deep into voice and instrument uniqueness. Violins sound sweet. Break-in time is about half complete.

I still consider this project to be experimental and not a finished design. Nevertheless, this project would benefit from better ears and other ideas if a couple of persons would like to join me in this adventure.

It's time to explain what I mean by a 16-bit "Sine wave weighted average" digital filter. Below is a picture of the XCEL spread sheet I used to calculate the resistor values to create a sine wave profile over 16 bits. I can send the actual spread sheet to anyone who PM's me their email address. I used two resistors in series to achieve an accuracy within 0.05% from theoretical.

The lowest resistance I used was 6800 ohms for the middle two positions. This is a little lower than the latest 8K recommended for the 32-bit Chinese equal weighted filter. There may be a better sounding resistor values. The equivalent combined resistance for 32 8K resistors is 250 ohms compared to 665 ohms for the 16 sine wave profile resistors. This may explain why the 32-bit filter sounded more forceful through the output transformers. I overcame this using a JFET buffer which am using right now and will stay with. I can share my buffer circuit, PCB, Eagle files, or whatever is needed.

If someone wants to try lower resistor values to try directly with output transformers without a buffer, I will do all I can to help. If the middle two resistor positions are 2500 ohms this would result in a combined resistance of about 243 ohms. Maybe 1000 ohms would be even better. I don't know and I am not going to find out.
 

Attachments

  • DSD 16-bit Resistor Values.JPG
    DSD 16-bit Resistor Values.JPG
    706.3 KB · Views: 803
Good news and bad news. First the bad news - my inner ears are full of sinus fluid so my hearing is a bit impaired at this time.

The good news is that this DSD no-DAC is very listenable. I can listen for hours with no fatigue. Bass is still deep and tight. No noise. Lots of musical energy. I can hear deep into voice and instrument uniqueness. Violins sound sweet. Break-in time is about half complete.

I still consider this project to be experimental and not a finished design. Nevertheless, this project would benefit from better ears and other ideas if a couple of persons would like to join me in this adventure.

It's time to explain what I mean by a 16-bit "Sine wave weighted average" digital filter. Below is a picture of the XCEL spread sheet I used to calculate the resistor values to create a sine wave profile over 16 bits. I can send the actual spread sheet to anyone who PM's me their email address. I used two resistors in series to achieve an accuracy within 0.05% from theoretical.

The lowest resistance I used was 6800 ohms for the middle two positions. This is a little lower than the latest 8K recommended for the 32-bit Chinese equal weighted filter. There may be a better sounding resistor values. The equivalent combined resistance for 32 8K resistors is 250 ohms compared to 665 ohms for the 16 sine wave profile resistors. This may explain why the 32-bit filter sounded more forceful through the output transformers. I overcame this using a JFET buffer which am using right now and will stay with. I can share my buffer circuit, PCB, Eagle files, or whatever is needed.

If someone wants to try lower resistor values to try directly with output transformers without a buffer, I will do all I can to help. If the middle two resistor positions are 2500 ohms this would result in a combined resistance of about 243 ohms. Maybe 1000 ohms would be even better. I don't know and I am not going to find out.

Carlsor,

This an inventive approach, 10 points for lateral thinking.

Having said this, I'm not sure you understand how a moving average filter works. Each resistor is responsible for passing the full signal, IOW it is not as if it was a ladder dac and the resistors are responsible for individual bit weights.

What you have proven however is that resistor matching is not important with this type of filter. I have simulated this on Ltspice and there was no difference in distortion with changing of resistor values.

What is important is that for each 'bit' or resistor, when it is on, the value is
always the same and same applies to when it is off.

As such this circuit will always work better into a virtual ground as opposed to a voltage OP scenario. Why? Because the logic gates OP impedance is part of that resistors value (probably somewhere from 10 to 20 ohms). If the OP is swinging in voltage, then the logic gates 'on' condition is not always the same, there will be a slight variance in current depending on where the OP voltage is. With a virtual ground I-V, 'on' always sees exactly the same load or current because the load resistor is grounded at I-V IP.

You can see this in, for example, Sabre ES9018 DAC's. They have lower distortion in 'current' OP mode than 'voltage' OP mode.

I know this may be difficult to understand but it is the way these circuits
work.

There are definite improvements that can be made to this circuit to make it more linear that involve the way the drive logic is implemented but as soon as I explain this here, we will then probably see another Chinese kit with it implemented :) :)

WRT the transformer and bass. Yes as you found, the transformer needs to be driven from a low impedance source for best results. This will make a big difference. Once you drive transformer from low impedance source then it's best to check it's transient response as it may be ringing and need a snubber. If the transformer is ringing, it won't sound quite right and you may even blame this on something else.

The best way to check this is drive transformer with a 10kHz square wave
simulating the existing circuit conditions and see what comes out the other end. The optimum condition is perfectly controlled vertical edge of square wave with no overshoot or ringing.

I like your enthusiasm - keep it up and good luck.

Terry
 
NO-DAC DSD Theory Discussion

I was hoping someone would discuss the theory, mathematics, etc. of what I have designed. Thank you, Terry. I tried to find the algorithm for converting an analog system into a DSD bit stream, but couldn't find anything. Because the DSD bit stream comes from a complex blend of musical sine waves my intuition told me to convert the bit stream into a sine wave segment. Also, RC filters act as a LPF on sine wave frequencies.

I would like to learn LTSpice simulation someday, however I was surprised to learn that it could do a distortion analysis of the analog signal coming from a matrix of resistors fed by bit shift digital chips. Is the distortion the same for 32 8K resistors as it is for 16 sine wave weighted resistors? They sure sound different to me. If the resistor values don't matter, then the 16-bit sine wave resistor profile should sound the same as 16 10.6K or 12 8K resistors which would have the same combined resistance.

My main concern was that small non-linearity's and lack of exact symmetry in the switching of the 74VHC595 IC might result in audible distortion. Does LTSpice take into account the imperfections of this chip compared to the ideal?

The 74VHC595 output internal resistor is 25 ohms which I took into consideration when calculating the external resistor values so that the combined resistance followed the desired profile. It shouldn't matter whether the resistance is inside or outside the chip.

I am treating this as a voltage output device. I understand how the AD844 is used in the DSC1 single-ended no-dac to create a constant voltage virtual ground to convert the 74VHC595 into a current output device. Perhaps the outputs of a differential pair of these can be used to drive the output transformers.

I have no way to inject a 10K square wave 2xDSD DoP signal into my setup. When I did this with my DDDAC using a PCM signal, the output of the Cinemags did not resemble a square wave even though my scope showed a square wave coming from the NOS PCM1794 DAC. I load the Cinemags with 10K-15K per manuf recommendation. I also put a 150pf WIMA across the outputs because it sounds better.

In the final analysis - how it sounds is all that matters, however, those of us infected by "audiophilea" are always looking for something better sounding.
 
Last edited: