• Please review our updated Terms and Rules here

Linear power supply fix - recommendations wanted!

1

1980s_john

Experienced Member
Joined
Apr 29, 2005
Messages
364
Location
UK
Hi,

I am currently fixing up a SWTPC 6800 computer which has a linear power supply (mains transformer, big capacitor, bridge rectifier). The system is designed so that the PSU provides 7V to 9V, and each card has one or two 5V regulators (Motorola version of 7805) to supply the ICs. I have got as far as cleaning the system, making the mains wiring safe (well safer anyway), and last weekend I powered up with a series light bulb and variac one board at a time. None of the boards had any shorts, and all the boards now have power. The problem is as follows. The system was built to UK spec for 1979 ie 240V AC. In the house where the system is the mains power measures 220V (the UK standard is now 230V +/- some small %, so this is within tolerance). However the power supply is struggling. I have removed the series light bulb and am driving it from a variac, which outputs 100% of the supply voltage to the computer, via a mains power meter. With all the boards in, the system takes 55W. However the voltage across the main smoothing capacitor is only 6.95V (was 8.0V with no boards), and on each of the board regulators there should be 5V. Most of the boards are OK (as they draw little power). My main worry is the two static RAM boards and the CPU board, these show 4.4V and 4.8V respectively, which I think is no good.

I have had a think and came up with four possible solutions, I would like to canvas opinion - the internet has many so this should be easy :)

Option 1 (my preferred). Replace the 7805 on the 3 boards affected with low voltage drop equivalents. These claim to give 700mV drop at 1A, so should easily provide a good 5V output with 6.9V input.

Option 2 - try increasing the input voltage to 240V (ie connect a transformer primary with 240V/220V tapping). I guess a 200VA transformer would be OK for 55W out. In theory this would increase main smoothing capacitor voltage to 7.5V and hopefully bring up all the regulator outputs to 5.0V. Disadvantage is cost & slightly more strain on the SWTPC transformer.

Option 3 - re-wire the system to 110V and drive via a variac. I assume output would raise to 7.5V as per option 2 (and using the variac I can always slightly over drive it eg to 115V). This has advantage of lower mains voltages in the case (which doesn't meet modern safety designs). Disadvantage is the inconvenience and risk of connecting 230V one day by mistake.

Option 4 - look at improving the wiring & bridge rectifier to increase efficiency. Least favourite option due to need to rewire, and supply is only drawing 55W so should be well within margins (actually another thought here - I don't know what the current drawn from the secondary is, if most of the 55W is drawn at 5V this implies 10A - wiring looks a bit thin for 10A so may be an issue, perhaps I should try and measure the current?)

Any views please!

Regards,
John
PS I have yet to connect up a terminal - like to check things thoroughly one step at a time!
 
SWTP was notorious for under-specifying their power supplies. The best thing would be to replace the transformer. But there are some other things you can do.

First, try replacing the rectifier with a schottky model. A standard silicon power diode will drop in excess of 1V per diode; schottky power diodes drop about half that. So, if it's in a bridge configuration, you'll gain an extra volt of output.

Another thing you can do is replace the board regulators (7805 type, I'm assuming) with LDO equivalents, such as the LM2940CT-5.0. They're drop-in replacements and require less than a volt of "headroom".

Neither of these measures is a wallet-buster.
 
Thanks for your reply Chuck. I've Googled for Schottky bridge rectifiers, but can't find one in a KBPC package (eg 20A) to match the original (also needs zero rewiring as it uses push on tags) - any links please?

For the low drop out regulator, I was looking at this one:

http://www.maplin.co.uk/1a-low-dropout-fixed-positive-voltage-regulator-46321

Other mod would be to add a 10uF tantlum capacitor across the output (needed for stability).

Regards,
John
 
I don't have anything in a KBPC using Schottky diodes--I think the best you can do in the way of integration is TO-220-style packaging. You could always design a small PCB with push-on-tabs if you wanted to stay as original as possible.

Most designs using the old 78xx style regulators already have (or should have, if you believe the datasheet) a decoupling cap on the output, so you probably don't need to add another one. Because you also probably have a pile of 0.1 uF ceramic bypass caps on the same line, there's no particular need for tantalum--aluminum electrolytic caps should work just fine.
 
I'm definitely with the LDO regulators, - least obvious "fix". Again, must ask the question.... what's the PSU ripple like?
 
Good question, I am not sure how much ripple there is (no oscilloscope was handy). I am fairly sure my DMM showed close to 0V AC across the main smoothing cap, but I can't remember testing it with the full load. BTW that capacitor is BIG. I spent several hours gradually building up the volts on it (series light bulb and variac), and with no load it holds a charge for many minutes.

I hadn't thought to try and measure for ripple on any of the regulator outputs, very good point - thanks.

Regards,
John
 
Last edited:
If there's not a significant (in the vicinity of 10%) ripple at the input of the 7805 regulators. you're not going to see much at the output of them--provided you have at least 3V of "headroom" on the supply.
 
Hi,

imho the first thing to do in this case is to swap the molded bridge rectifier for a bridge made of standard discrete silicon power diodes.

I have seen hundred of these molded horrors exhibiting large voltage drop across them due to an abnormally high internal resistance, sometimes even when they are new. They seem to be often produced in poorly equiped Chinese kitchens using batches of out of specs diodes rejected by components manufacturers.

This become worst when these devices are ageing. They are cheapo components which must be avoided in every serious design or at least they often need to be changed after a few thousand hours of operation even if they seems electrically ok.

I don't believe there is any benefit to use low drop regulators as there is nothing wrong with the original design of the SWTPC. Just supply the existing circuitry with a good rectifier and filter cap.

Hope this helps
 
Good point, Thierry! But given the age of the system, the KPBC bridge was probably produced in an US kitchen. China had only just been visited by Nixon and trade relations were a pipe dream.

Does the KBPC bridge run hot?
 
Last edited:
Yeah, I'd check the bridge as well before making "design changes" ;-) ; maybe one leg is even open?

8V no-load sounds kinda low to me; what's the no-load and full-load AC in? (to the bridge, i.e. the transformer output)
 
Last edited:
You're absolutely right Chuck(G), apologies for this mistake about the nationality of the kitchens cooking bridges in the era!

Good question MikeS.

Ps using 7805 series reg needs about 9.5 V. A.C. at the transformer output resulting in about 13VDC on the filter cap with absolutely NO load.

AC as well as unregulated DC will drop with load but must it stay above says 8V. dc. 6.95V. are way too low.
 
My recollection of the TV Typewriter kit from this time was that the power supply was somewhat under-rated (those 2102s sucked a lot of power!). I ended up re-designing the power supply and it ran fine until I retired the unit.

It could well be that the SWTP kit for the Eurozone was equally dicey. MITS was also terrible with their power transformer specification on the 8800.
 
Hi,

Update on the SWTPC 6800 power supply. I went away and bought some LDO 5V regulators from Maplin, and fitted two (so far) to two of the memory cards where one of the normal regulators was struggling (these cards draw so much that they have two regulators). Output is now 5.0V. However, the mains input was a bit better at 230V, and main capacitor voltage was 7.5V, so may still be sensitive to slight drops in mains voltage.

I borrowed an oscilloscope, this showed 400mV p-p ripple, which I think is quite good. Maybe that big capacitor takes 10s of hours to fully stabilise.

Anyway, I've had to remove the memory boards for some other testing, so thanks for the help on the PSU, I think LDO regulators will be the easiest fix if problems with 5V supplies occur again on this machine.

Regards,
John
 
You must log in or register to reply here.
Back
Top