Replacing The STV-3H and -4H Diodes

EchoWars

Hiding in Honduras
Sponsor
Seems these are getting rarer and rarer, and I was even PM'd a few days ago by a needy AK'er who told me a source wanted $50 each for some STV-3H diodes. Screw that (and the seller needs a good kick in the nuts, if you ask me).

Regardless, most of the time these are used in biasing circuits, where a few milliamps of current is sourced through the diode and used to set up the voltage between the bases of the driver transistors. Therefore, the most important specification is the forward voltage. This varies according to current, so in an attempt to duplicate the same voltage as the original devices I measured them at various currents between 500µA and 20mA. I then used combinations of diodes to emulate the originals. The diodes used were 1N4148, UF1004, the 1N3595, and the BAT41 Schottky. The resulting chart is here (click to enlarge):
diodecompare4_zpsiiathvha.jpg~original


I have arranged the devices used to match their location on the chart for ease of interpretation (top graph is for the STV-4H, next one down is the 1N3595x3, etc, etc...)

Let's look at the results with an eye towards replacement. The top graph (magenta square) is very well matched with a series string of three 1N3595 diodes (hollow green square). Simply can't do much better than that. The string of three 1N4148 diodes (electric green triangle) could work in a pinch, but the voltage values are consistently 100~120mV low. This should be no problem for the bias adjust to compensate for.

I did string four 1N4148 diodes to see their Vf, but the resulting voltage was much too high, and I omitted plotting it on the graph as unusable.

The next graph down (burgundy square) is a series string of two 1N4148's and a single BAT41 Schottky diode. The Schottky is reputed to have a very low Vf, and I had hoped that this combo might be useful to replace the STV-3H diode. As you can see, the Vf of this diode string is within spitting distance of the STV-3H until we get beyond 2mA, at which point the Vf of the BAT41 begins to rise very quickly, and approaches that of the three 1N4148 diodes. I did not test beyond 20mA, but I feel that above that point this diode combination probably would have exhibited a larger Vf than the three 1N4148 diodes. I did further testing with more BAT41 devices, and found the same thing...at very small currents, the Vf of the Schottky's is quite low, but rising above 2mA the Vf of these devices increases quickly, so that as we approach 20mA we have nearly equaled the 1N4148 silicon devices.

The next graph down (blue diamond) is the graph for the STV-3H. You can see that it is closely shadowed by the three UF1004 diodes (brown circle), which have a nice linear Vf up until the last, where Vf begins to increase a bit more rapidly and crosses the STV-3H graph at the 10mA point. This looks to the the best replacement for the STV-3H...but is a little unfortunate, since the size of the device and the large leads will make this a cumbersome swap. The next graph down (hollow blue circle) is two 1N3595's, which is consistently 225mV or so low as compared to the STV-3H. Reasonably close, and the bias adjust could make up for it, but with only two diode junctions I fear this would not work well, as the end result would probably not compensate well for temperature variations (which is the whole point of using these devices).

But...there are thousands of diodes out there, and I hope to perhaps find a small surface mount device that gives an approximation of the voltage needed for the STV-3H diode. And this chart will help if you also decide to start testing for possible replacements, since here I have given the Vf of the two most common STV diodes at various currents.

Good luck gentlemen.

1/2/12 Update: I recently had a Spec 2 here with a missing STV-4H, and had to come up with a replacement. So another device was built with three 1N3595's, and since I had it in-hand, I decided to check temperature tracking by warming up a small glass full of cooking oil and comparing the three 3595's to an original V-4H. What I found was a little disappointing, as the 3595 did a poor job of following the V-4H as temperature rose...it simply did not drop its Vf enough to make me happy that it is a good substitute. I did use it on the Spec 2, however, since despite the V-4H being heatsink mounted on this amp, it is used only as a failsafe device and not used for bias tracking.

I did build another device out of four UF1004's, and did another temperature check as well as a current check, and found that four of these devices follow the V-4H pretty well.

So, thus far, the only devices that I have found that mimic the originals reasonably well is the UF1004 (three UF1004's for an STV-3H and four UF1004's for a STV-4H), which is a bit of a bummer, as the device size and the lead thickness make it an unwieldy replacement. What I did do, to minimize the lead diameter issue, is to clip the original leads and use small leads cut from 1/4W resistors to join adjacent diodes. That at least bypassed the lead size issue...if only the body of the UF1004 was a bit smaller.
 
Last edited:
That is a very useful and informative piece of work Glenn.

I have been thinking about this very subject recently as there seem to be several different types of these used in my newly acquired AU-D11 II. Luckily they all seem to be good and the amp sounds absolutely marvellous, but you never know when the information you have posted will come in handy. :thmbsp:

Thank you. :thmbsp:
 
This is very good stuff! Thanks. You know the day is coming when the original parts are NLA. Did partstore run out of stv3h? A bought a couple from them recently. They were about $5 each.

- Pete
 
Nice work, Glenn! Have you come up with a good way to mount the axial subs to the heatsink?
 
Thanks. Great work.

Agree wholeheartely about the kick in the nether region!!!!!!!!!!!!!!!!

Greed of people just amazes me at times.
Don't know when it became ok to try to get over on others but seems to be the norm today.
 
Nice work, Glenn! Have you come up with a good way to mount the axial subs to the heatsink?
I got ideas...a ring terminal with the diodes affixed to the crimp section (opened up and flattened) with JB Weld. Haven't tried it yet.
 
Well, whoever tries it first can post results here. I have not replaced one of the STV's in a few years, for some reason. But now I know what to do.
 
The perspective may change depending on whether you are a parts seller, or a repair technician, or combination of the two. If you sell parts for a living, any condemnation is deserved. If you're a repair tech who is not in the business of selling parts, then this may fall under the category of 'anything is for sale, if the price is right'. If the latter is the case, then my first post is probably unfair. I have a couple of each of these diodes, and given their rarity they are not for sale at any price. For me, it's easier to say that, than to specify some outrageous sum at which I'd part with them.

And for those who in the future may beg for me to sell them one of these diodes, I can point to this chart and say that, with a little creativity, you don't need my diodes. And I don't have to quote that outrageous price where I'd let them go. ;)
 
One thing to watch out for is over temperature how well the home made diode string matches the original. The temp coeff should match the transistors to keep the output stage bias stable over temp.
 
It'll be approximate, and hopefully somewhere in the ballpark. That's the best we can do, unless you've got something up yer sleeve.
 
Count the number of transistors the string is biasing. They should be Si transistors in all cases now. That is the number of Si diodes to use. I believe the STV-3 is used with the same number of trans in amps. Si diode and Si base-emitter junctions have the same temp coeff. Adjust the current through the string to get the proper quiescent current for the output stage. The diode forward bias voltage can be varied with current which will determine how much the output Q current is at. Some amps has a resistor in series with the string to keep the diode current small. When Vbe changes due to temp, the diode string should keep up with the change.
 
What should be apparent is that I know all of this. ;) I thought you might have had a rabbit to pull out of a hat or something... Oh well.
 
It'll be approximate, and hopefully somewhere in the ballpark. That's the best we can do, unless you've got something up yer sleeve.

Well, the best we can do is probably to use a transistor with a voltage divider in the base circuit (or resistor and pot) to set the bias. Then you only have the transistor to clamp to the heat sink, and it probably emulates the output transistors Vbe curve better.
 
The problem with that is the heatsink may be a long ways away from the driver board. Long leads on this transistor may cause stability and noise problems. On other amps, because of shielding around the driver board, it isn't possible at all (700M comes to mind). But, in some individual cases it's probably a workable solution. However, it's a lot easier to simply pop in another diode, which brings us back to the chart in Post #1. ;)
 
>I know all of this

Then the numerical part number is not critical, the real work is figuring how to mount the diode chain for good thermal conduction/tracking from output transistor.
 
>probably to use a transistor

It is more than the output trans in some cases . You need to count the number of b-e juctions the diodes are biasing. We need to compensate for the sum total b-e.
 
I got ideas...a ring terminal with the diodes affixed to the crimp section (opened up and flattened) with JB Weld. Haven't tried it yet.

I was just inventorying bias diodes because of this thread, and ran across a couple of unidentified parts made just like you suggest. That's exactly what I would do, if I were makin' em.
 
Back
Top Bottom