• Please review our updated Terms and Rules here

Tarbell Floppy controller modification

new_castle_j

Experienced Member
Joined
Dec 3, 2011
Messages
318
Location
Texas, USA
What is the purpose of this giant ceramic resistor on the back of my Tarbell floppy controller? Obviously it's not stock.

IMG_4160.jpg IMG_4161.jpg
 
It's a shunt around the series regulator (+5) to improve the current handling capacity at the expense of regulation accuracy. If you don't like it, you could replace the TO-3 regulator and resistor with a switching regulator, which would run much cooler. You saw this sort of thing on early MITS boards with 7805 regulators. You can see from the discoloration on the PCB that the regulator is running very hot.
 
Under estimating the thermal load seems to be quite common on these early S100 boards. The +8V line is often around +10V adding significant thermal load on the regulator. The heat sink on this board is even small for a TO-220 regulator. For a TO-3 running anywhere near its rated current it is a miss. Most of these engineers know Ohm's law. It is not rocket science to get the thermal resistance of these heatsinks.
The temperature that this board was running at was close to melting solder.
Dwight
 
Is there any particular reason to keep the TO-3 appearance? LM2596 "simple switcher" based regulator modules are pretty cheap and small. 3A output.

Figure it this way, if your supply is +10 and you're drawing 2A at 5V, that regulator will dissipate 10W, which is a lot even for a PCB-mounted TO3 with a too-small heatsink. A buck regulator won't break a sweat at that power.
 
Last edited:
Is there any particular reason to keep the TO-3 appearance? LM2596 "simple switcher" based regulator modules are pretty cheap and small. 3A output.

Other than having the pinouts to the board the same, no.
I ended up just buying some cheap modules and will make a board with the TO-3 pinouts and solder them onto it.
 
Under estimating the thermal load seems to be quite common on these early S100 boards. The +8V line is often around +10V
Dwight

It is because of this I run my SOL-20 from a Variac and set it so it is around 8V, if it gets too close to 7.5V it reaches the 2.5V in/out difference required by the standard 7805 and ripple appears in the supply output. Over 8V to 8.5V its not good as the heat dissipation rises significantly. I installed a small volt meter in there, but I found that at the point that the input voltage is too low, close to 7.5V, ripple appears in the video out that is easy to see on the VDU, so I just increase the line voltage until it just goes away and a little more and I don't have to have the top off the computer to see the voltmeter.

Some of the early S-100 memory cards were very power hungry too with multiple regulators to share the load.
 
Last edited:
One thing to realize is that a regulator in a TO-3 can will get to the same temperature ( all other things equal ) if it is rated at 1.5 amps or 3 amps, with less than 1.5 amps. So, putting an over rated regulator is just costing more money that could be invested in a better heat sink. In fact for a S100 card, you are more likely better off putting two regulators on the board with the 1.5 amp rating than one 3 amp one. Another point is that it is better to put them on different ends of the board because the slight drop across the board will help them share the current load better( some RAM board designers never got this ).
Also, check the heat sink's designer recommendations. Changes in orientation can make a difference. The one used on this board was likely the worst from what I recall but it is only a small difference.
Dwight
 
You have to wonder what's drawing the current. I have the original Tarbell 1771 floppy board and it uses regular 78xx TO-220 regulators that don't get overly hot. I don't remember that 8257 DMAC getting really hot on my other boards.
 
Well, if there were a demand, it wouldn't be difficult to make a TO-3 shaped PCB. Maybe glitch wants to do some--not my thing.

Sure, we can run an adapter board if folks are interested in that! Those EzSBC modules are usually pretty well built though, I don't think I could probably run a similar module for less, not fully assembled + tested.

You have to wonder what's drawing the current. I have the original Tarbell 1771 floppy board and it uses regular 78xx TO-220 regulators that don't get overly hot. I don't remember that 8257 DMAC getting really hot on my other boards.

Could just be really poor/zero airflow. I have a bunch of Vector Graphic ZCB boards from embedded applications where basically they were stuck in a box with no airflow, they're all very discolored around the regulators, even though that's not a board that runs hot.
 
I don't think I could probably run a similar module for less, not fully assembled + tested.

Thats why I was thinking of just an adapter board for the 50 cent Chinese ones, like
https://www.aliexpress.com/item/32796268715.html

I don't know if they are small enough to fit between the TO3 holes yet, though.

Grabbed a picture, looks like they are around .3 x .8" TO-3 hole spacing is 1.8"
so they would fit easily

Screen Shot 2019-08-24 at 9.40.44 AM.jpg

A hexagon would even work for the pattern, and it would panelize easier
 
Last edited:
Yup, the same ones are available on Amazon and eBay in bulk for about a buck apiece, if you don't like waiting for shipment from China. There seems to be a choice between one with a trimpot for voltage adjustment and one with a fixed resistor divider.
 
. Changes in orientation can make a difference. The one used on this board was likely the worst from what I recall but it is only a small difference.
Dwight

I think S-100 boards do a lot better when they are mounted vertically because of the convection air currents. In the case of the SOL-20, with the boards horizontally, heat building up between them is a real problem, especially for the ones closer to the top of the pile. So I added a small fan to blow air between them to help keep them cool and it made a big difference along with keeping the pre-regulator voltages as low as possible.
 
The fan blowing through the card stack was a standard option for multibus cages.

One of the reasons that I don't like the PC system of expansion cards. The NEC c-bus was better in this respect as the card contacts were opposite the mounting bracket, so air could be blown through the stack.

Too bad we're stuck the system now.

04131638_4da552f25da49.jpg
 
Tthe card contacts were opposite the mounting bracket, so air could be blown through the stack.

The PCB layout also flows better. Looking at the NEC card all the bus buffering is to the left of the Ethernet IC.
Apple's implementation of Nubus copied the same mistake putting the DIN connector next to the output fence.
IBM didn't make that mistake with Microchannel, but then PCI moved it back.
Note that it is not necessary to open the box to insert the NEC card, and you can use a lever to insert/remove cards.
You don't have short length cards, so there is a card guide on both edges.
VME, and HP's DIO, even S-100 are other examples of doing it right.
 
Back
Top