VGA/TV DAC using resistor arrays (and other uses)

Peter Jakacki

A long long time ago, somewhere in the deep recesses of the Propeller forum I posted a diagram of the arrangement that I use to create the video DACs using resistor arrays. Theres are normally made up of discrete resistors of various values that "should" be binary multiples of each other. I find that smd resistor arrays are costing me about half a cent each and they pack 4 resistors of the same value in the same "space" that a single 0805 resistor takes. At the time I implemented my DACs for the TV or VGA using the same pins and just used 470R resnets which seems to work quite well.

Here's what they look like (look to the bottom of the RJ45 socket):

For the purists however I present a similar implementation using three 240R resnets plus a single (or two) 560R resnet(s). Here's a schematic:

So why do it this way then? Well, the advantage is that you can just use the one or two values of resnets rather than all the individual values of resistors. It's also very compact as you can see from the photo. Plus this is a true binary DAC as the values increase in powers of 2.

There is a slight mismatch on the TV DAC which can be easily corrected if you want by just using another 560R resnet to get 280R or just put in a 33R resistor in series.

Is there any other advantage? Yes, using resnets in place of discrete resistors means that you can stock less values of resistors and instead make up values from combinations of the resistors in the resnet in the same space. Here's an example where we use a 240R resnet to get these odd values:

1) 60R using 4 in parallel
2) 80R using 3 in parallel with 1 spare
3) 120R using 2 in parallel with 2 spare
4) 240R using single elements with 3 spare
5) 320R using 1 in series with 3 in parallel
6) 360R using 1 in series with 2 in parallel with 1 spare
7) 480R using 2 in series with 2 spare
8) 600R using 2 in series with 2 in parallel
9) 720R using 3 in series with 1 spare
10) 960R using all 4 in series

So you can imagine that with as little as 3 values of resnets that you can span a wide range of values. Digikey sells reels of 5,000 for $25, that's 20,000 resistors. So that's my little design tip for the day.

Peter Jakacki

Seeing nobody made any comments on this one I thought I would throw in a little bit more about using resnets in place of discrete resistors. I wrote a little program that spits out the values from all practical combinations not counting any spare elements left over. I chose values in the hundreds just to give an idea of the range of values obtainable. Did I leave any combinations out? Did I make any mistakes?

Here are a few common values with uncommon results:
(2S+2P means 2 in series plus in series with 2 in parallel)

4 ELEMENT RESNET COMBINATIONS 
  
RESNET  100
 4P           25
 3P           33
 2P           50
 1S           100
 2S           200
 3S           300
 4S           400
 1S+2P        150
 1S+3P        133
 2S+2P        250
  
RESNET  220
 4P           55
 3P           73
 2P           110
 1S           220
 2S           440
 3S           660
 4S           880
 1S+2P        330
 1S+3P        293
 2S+2P        550
  
RESNET  240
 4P           60
 3P           80
 2P           120
 1S           240
 2S           480
 3S           720
 4S           960
 1S+2P        360
 1S+3P        320
 2S+2P        600
  
RESNET  330
 4P           82
 3P           110
 2P           165
 1S           330
 2S           660
 3S           990
 4S           1320
 1S+2P        495
 1S+3P        440
 2S+2P        825
  
RESNET  470
 4P           117
 3P           156
 2P           235
 1S           470
 2S           940
 3S           1410
 4S           1880
 1S+2P        705
 1S+3P        626
 2S+2P        1175
  
RESNET  560
 4P           140
 3P           186
 2P           280
 1S           560
 2S           1120
 3S           1680
 4S           2240
 1S+2P        840
 1S+3P        746
 2S+2P        1400
  
RESNET  680
 4P           170
 3P           226
 2P           340
 1S           680
 2S           1360
 3S           2040
 4S           2720
 1S+2P        1020
 1S+3P        906
 2S+2P        1700

Tubular

Hi Peter

Good work. Nice to have someone pre-think these kind of things through. I think Phil Pilgrim once posted some info about the 430R resnets used on the propeller backpack.

I've been doing some work with 64 greyscale VGA, and one thing I noticed is the output resistance of the prop pin should be taken into account for small values of external R. I did some brief calcs and estimated the resistance at 27 ohms at 3v3, but haven't had a chance to characterise it properly (high/low logic level, Vdd supply voltage, temperature, shunt reactance, variation from prop to prop).

The good news about the 27 ohms is some white and blue leds can be directly driven, albeit with not great intensity matching, temperature independence, etc

Peter Jakacki

Thanks Tubular, I think my earlier posts predate even Phil's

The 27ohms source resistance you are seeing is probably close to the mark but since this is the MOSFET channel resistance when on I don't think you can quite count on any particular value like we can with fixed 1% resistors!

There is a trick to driving LEDs directly from a higher voltage that I have used in designs. Seeing that the LED has a forward voltage drop means you can drive your normal LEDs directly from the Prop I/O lines (through resistors) even if the LEDs are tied to +5V. So much more is the case for LEDs with higher voltage drops such as white and blue. If you have an LED with a 3V drop then you can certainly tie this to +5V and calculate your current limit based on 2V across the resistor.

TimH

Another advantage of using resnets in place of discrete resistors is the reduction of the thermal coefficient ratio. Being packaged together the resistors see the same temp changes. When using the resnets as voltage dividers, the output is much more stable.

ericball

Probably a silly question, but with the circuit in your schematic aren't you going to get crosstalk between HSync, Vsync and B via the TV output?

Cluso99

I was going to ask the same question as eric. I see it forms a voltage divider of 1120R : 240R which is probably ok.

BTW from what I have asked, seems the 4th pin for TV is not used often, if at all. I have commandeered this pin for other uses.

I have pondered the connection of VGA/TV using sommon resistors. I was wondering if the VGA would suffer if 270R & 540R were used instead of 240R & 470R, or alternately would TV suffer if 240R, 480R & 960R were used instead??? This would make a simpler network.

The only issue with resnets is they are harder to hand solder due to the fine pitch. They are the most difficult part on my RamBlade.

Peter Jakacki

@ericball: Crosstalk? This circuit is used for either VGA or TV but not both at the same time. Any signals that appear to be cross-coupled are terminating back into low-impedance source pins (the Prop's I/O) and so are "swamped" from causing any interference. Sketch the paths (include the source/sink resistance of the Prop pins as well) and do the maths, you can see there is no problem.

@Cluso99: The VGA resistors will only affect the luminance of each of the primary colors and there is enough difference from monitor to monitor anyway so it shouldn't matter. The series resistors in the sync lines aren't really necessary but does protect the Prop. I'm using 470R resnets exclusively for both the VGA and TV and I can tell you that it just works.

Cluso99

Thanks Peter.

Cluso99

Pondering a little further...

I can get this to 3 resnets, 2x 240R and 1x 470R. I think this would work. You will note I include an optional terminating resistor for TV of 191R-200R.

I have marked 3 links, but I think they could be permanently connected without problems. Just select the correct output in software.