Here you see the well known DDDAC Power supply. I designed it to go together with the DDAC1794 kits, Of course it can also be use for many other audio applications.
As DDDAC kit it comes as “5V” or “12V” version. As it is adjustable, the range is a bit wider than that actually.
The kit is laid out to be a 1A max power supply. But is that really the max you can get out of this? And… are there other tweaks? How does this PSU actually work? Often heard: can I make a 2A or 3A PSU out of it?
This blog will try to answer those questions and will give some good guidance of the electronic background and deliver a few tweaks for everyone who wants to beef it up or experiment with the sound quality….
This is how it basically works:
As you can see, it is a simple principle without rocket science. I use full wave rectifying with schottky diodes. After that comes the filter to make a raw DC voltage out of it with some ripple on top (which is a sawtooth 100Hz, which has nasty harmonics). Here is a simplified RC rectifying filter shown. In reality it is a “RCRC” filter, but later on I will show a “RCRLC” solution which does more filtering.
To arrive at the desired output voltage, there is a discrete error amplifier which compares the output voltage with a voltage reference (the green diode) and drives the output darlington. Key here is to also filter the supply voltage for the error amplifier. As any ripple and noise left over will find its way, even when attenuated, into the output.
The error amplifier in detail:
The supply voltage for this stage is filtered by R11 (330 Ohm) and C11 (470uF). This will result in a 1Hz corner frequency. Resulting in a -40dB pre-ripple rejection at the 100Hz first harmonic. An extra benefit is that this supply DC voltage does not have the larger saw tooth. For example at full load the RAW DC ripple can be peak tot peak 1 Volt or so. In the error amplifier this will be only 10mV. The supply current is roughly 5mA so you loose 1,5 volt here. This is a compromise between amount of filtering and losing voltage control headroom – later more on this – You can also increase C11 but somehow I could not measure a huge impact. Nice is the slow start of the PSU by doing this. But I leave it with 470uF to be honest.
The actual error amplifier is T5 and drives T6 which in turn drives the output darlington with low impedance (emitter follower). T4 with the green LED is a current source, to maximize the gain of the error amplifier. All very straight forward and this setup results in a ripple and noise rejection (PSRR) of roughly -71dB over a bandwidth of 500kHz, which is a pretty good figure 🙂
Complete circuit with all values:
Maybe you noticed. Just through the backdoor, I introduced the first tweak…. Replacing the second RC rectifier filter section with a resistor and an inductor for additional ripple and noise filtering. Just by the way, mentioning, I supply this level of detail for your DIY purpose only. A ~ 1:1 copy for commercial use is prohibited. (that includes changing R11 to for example 270 Ohm 😉 ). If you are interested in any DDDAC design in your OEM product, just contact me.
Later on more on the choke / coil / inductor (what is it really?)
Let’s start tweaking
Impact of transformer size on headroom and power output
I guess this is the most requested tweak. How can I beef up the PSU to deliver more power? First you need to understand what the minimum required headroom of the PSU is…. Meaning what RAW DC input voltage do I need to still have a regulated output?
Good question: let’s calculate and MEASURE it 🙂
But first look at the design….. If you start with your desired output voltage. Let say 12,5 Volt for example. To be able to actually arrive at this you could work your way “up” throught the circuit:
- V-drop darlington (1,5V)
- V-drop T6 (0,7V)
- V-drop Vce T4+R10 (??V)
- V-drop Filter R11 (1,5V)
So this adds up to 3,7 Volt + Vce (T4) and R10. The last one is a bit variable depending on the load and control, but lets say ~ 1,3 Volt. In order for T4 to NOT saturate (the error amplifier will lose all gain and stop rejecting ripple…) you need at least 0,5 V to be at the safe side. With low load and hence high RAW DC Voltage, this is where the headroom sticks and it varies between say 5 volts and saturation in overload conditions.
In conclusion: 3,7 +1,3 +0,5 is 5,5 Volt – So we wanted in our example 12,5 Volt, so we need a minimum DC input (= average DC level of rectified and filter voltage) of 12,5 + 5,5 = ~18 Volt….
Do we have that available in the standard version?
Now transformers all behave the same: at no load they are having a much higher AC voltage output. This can range from 10 to 30% easily. On the other hand the secondary windings will have a loss in their Ohm Resistance of the winding. And … at this point everything will collapse… using the transformer above its VA power rating, the core will saturate and the voltage will drop rapidly. (and the transformer will heat up and burn out eventually)
I hope it make sense when I state that using a transformer way below it specs will keep the raw DV voltage more constant.
At the left you see measurements of the DDDAC PSU with the standard 25VA blue Talema transformer versus a RKT 120VA version
The top lines show clearly what I just described and also that indeed 1A is the absolute max you draw from the standard version. The bottom line shows the Vce over the current source. At 0,1 Volt it will saturate. Oh, that means with the RKT transformer you can go up to easily 2A? Yes that works fine… May be you ask, why not a transformer with higher AC voltage? Yes, will work of course but then you will dissipate much more power in the TIP122 output darlington. Can be done, but read further below what to do…
Before I forget… I indeed could confirm by measurement of the PSRR which dropped sharply from ~-70dB to -40dB when the delta between input voltage and output voltage becomes less than ~5,6 Volt…
The most important point in increasing the output power capabilities is the TIP122 power loss dissipation and the resulting heatsink dimensions….
So assume we use the 120VA (still 2x15Vac) and load the 12,5V PSU with 2A. We now have (measured) ~7 Volt delta. In the darlington we need to dissipate 2A x 7V = 14 watt – for the TIP122 itself no issue (datasheet: MAX: Ic=5A and Pd=65Watt) as the Rth case is ~2K/watt, the junction will heat up 30 K. Meaning the junction will be like max 100 degrees when we assume heatsink of max 70 degrees (really worst case this should be 😉 ) This de-rates the TIP122 to ~45% of 65 Watt = 29 Watt – so we are still in a very safe area with 14 watt as long as we stay below 70 degrees heatsink.
assume the inside of the DAC case will be 40 degrees (warm days not so optimal venting…. than we should have max 30 degrees of temperature raise in the heatsink to keep it below 70K. I would go to be extra safe for only 20 degrees more…. so we look at 20 degrees with 14 Watt for a safe case -> so Rth (heatsink) = temperature / power dissipation = 20/14 = 1,4K/W. That is quite a big boy…. can we do smaller, yes, lose the extra buffer I used:
So here is an absolute minimum case. This would be a raise of the heatsink with 40 degrees and a cooler case of 30 degrees and accept the heatsink to get to 70 degrees. That would halve the thermal resistance of the heatsink Rth= 40K/14Watt = ~ 3 K/Watt – quite doable… Still I would suggest a 2K/W…
Can the power supply do 3A maybe ? Sure no problem. You know now that you need a BIG transformer. Take 200VA or so and you also understood that you will dissipate ~ 21 Watt. So take a heatsink of 1,5 to 2K/W…. (change the fuse also 😉 )
Extra filtering on the raw voltage
A picture detail of how to easily do this tweak. I just soldered the combination of the Rseries (1 Ohm) and a 10mH Monacor coil at the position of the fuse. With this there is less danger of large damage at short period overload, So I left it out
Why 1Ohm and 10mH? Let’s have a look on the impact of adding a choke into the filtering of the RAW DC voltage
I am using TINA Spice simulation for experiments like this
For me the most important feature are the nested simulations, so you can “try out” the impact of a range of component values. At the right you see we need to dampen the resonance. I used a 1 Ohm wire wound resistor to minimize power dissipation and still have a relatively flat curve
Now I changed the value of the Coil / Choke inductance, still using the 1 Ohm series resistor. As you can see the 10mH Choke adds like -13dB rejection at 100Hz. With 1H it is even MUCH better. But a version with max 0,5 Ohm or so will be hard to find and/or expensive
In terms of sound quality…
me and also several others confirmed that using larger transformers and adding chokes is improving overall sound quality. Feel free to experiment and see for your self what it does for you. After all it is a “DIY HOBBY”, right ?
- PSRR is pretty good! 500kHz -71dB
- In case you need a higher output voltage make sure the RAW DC voltage after the RCRC filter is > Output-Voltage + 5,5 Volt (use 6 Volt to be “safe”)
- If you go beyond raw DC of 25V you need to change the capacitors (they are 25 V versions)
- You can easily increase the output power of the DDDAC PSU by using bigger transformers and bigger heatsinks (you could even parallel TIP122’s – to do like 5A – but I did not cover doubling in this post)
- It is advised to increase the first Capacitor to 10.000uF or 22.000uF when going to BIG loads – but NOT needed (!)
- Adding a choke increases the PSRR with -10 to -15dB depending on load
- Increasing the choke inductance value would help “more” – but be aware of the fact the increased resistance need to be taken into account !! (voltage drop AND power dissipation) Also it gets bulky and expensive 😉 Up to the DIY…
- Added later: Adding a choke and extra resistance will ad extra voltage drop – so “Choke” and “2A” were not combined in this Post (!) You might require a larger transformer or a choke+R with lower series resistance…
- I hope you enjoyed reading this !
If I was unclear or forgot to explain something, please use the feedback function below…
If you want to get an short email notice when my blog is updated, enter it at the right bar (top/mid somewhere…)