Issue 2 TR4 keeps blowing and DRAM is tested OK

[rimmeruk]

Hello all,

Just finished a repair (almost) of an issue 2 board 48K and having an issue now where TR4 transistor keeps overheating from excessive current.
I am using a bench power supply and if I limit the current to less than 1A the little blighter starts up fine. The machine runs fine
with a 1A current limit on my bench supply and only uses around 700mA max. All upper/lower RAM, ULA, ROM are passing tests fine with Retroleum's diagnostic interface (a superb thing I must say). I can start games and run the machine for hours with the bench PSU current limit.

However, if I turn off my current limit or used a 9V wall power adapter rated at 2A, I see on my multimeter the current
reaching nearly 2A when I start the machine up. The TR4 ZTX651 transistor then fries itself and I've wasted 8 of them now trying to
fix the board

So far, I have replaced all caps, 5V regulator and the TR5 transistor which does not have any issues. Does anyone have any idea
what could be causing this or what other components on the board could cause TR4 to keep burning up ? Like I said, the DRAM is
tested and working fine and I've gone over all the diodes with a multimeter, can't find any shorts on them.

There is obviously a short or partial short somewhere on the board that's causing the excessive current draw when powering on.
Just not sure where to start removing components. The ULA and Z80 have been placed in a good working donor board and operate normally.

If I can get it fixed I'll post some photos of the mod's I've done, otherwise I'm stripping the board as the issue 2's seem to have more power
related faults than any of the other issue boards. The issue 3 seem to be much more stable and reliable board to work on.

Any help appreciated and thanks for reading.

[1024MAK]

Has the DC-DC converter circuitry (TR4, TR5 and the coil and associated circuitry) been upgraded to Sinclair’s latest mod state? Details here.

Assuming that there is no fault with the DC-DC converter circuitry, TR4 typically dies if too much current is being drawn from the +12V supply. With no expansions fitted, the +12V supply is only used by the colour encoder chip (IC1, LM1889) and the eight 4116 (or equivalent) DRAM chips (IC6 to IC13). It’s rare for the LM1889 to fail. If you do suspect it, don’t immediately change it. Just lift one end of resistor R62 to isolate the +12V supply to it.

Is the -5V supply voltage okay? If the -5V supply is outside the range -4V to -5.5V, it can cause problems with the 4116 DRAM chips. The -5V supply provides the bias voltage to the semiconductor chip, without this, it can cause the chips to malfunction due to transistor junctions conducting when they should not, and can lead to the chips self destructing. Also are any of the 4116 chips running very hot?

Also it is worthwhile checking that all the electrolytic capacitors are fitted with the correct polarity. Don’t just rely on the silkscreen markings, double check by seeing visually or by resistance tests which tracks and components each capacitor lead connects to.

Mark

[rimmeruk]

Thanks for reply Mark. I've checked the link you posted to upgrade the DC-DC circuit as that is something I have not looked at.
The caps are all correctly positioned and yes the polarity is correct as I read up on the silk screen errors before the re-cap.
I have good +12V and -5V power lines, although not when TR4 blows, I loose these and get the obvious DRAM errors on boot.
Nothing is obviously hot apart from the ULA which seems to be too hot for my liking, burns my finger when touched. I've had it
running for more than an hour like this though.

When TR4 does blow and I have the diagnostics interface plugged into the expansion port, it reports IC1,IC2,IC3 failure.
Not sure if that's just a common error message for the missing +12v and -5V to the IC's.
I guess I will go over the board much more in detail with my meter and start lifting IC pins.
Would a faulty coil cause this over current ? Not sure how common they fail but it is the major component in the DC-DC circuit.

Thanks.

[rimmeruk]

It’s rare for the LM1889 to fail. If you do suspect it, don’t immediately change it. Just lift one end of resistor R62 to isolate the +12V supply to it.

Is the -5V supply voltage okay? Also are any of the 4116 chips running very hot?

Mark

I lifted the R62 and no change apart from no video output.
The DRAM chips are all cool.
The voltages on the DRAM chips are -4.85V +12.15V +5.03V.

I've also added the DC-DC circuit mod from Retroleums website (replacing R60 with 270R and adding the 4.7uF between TR5 and C34)
Still getting a huge current draw (> 1.5A) when applying power and now blown another 2 ZTX651's. These transistors aren't cheap
It will run for hours through my bench supply with a current limit of 1A.

Strange thing is that the excessive current draw is only when first applying power, after powering on I can disable the current limit and the speccy only draws around 750mA ?

I'm now thinking of adding a LM317 as a current limit to the board so I can run it from a 9V power adapter.
The spectrum operates fine with around 750mA and with the Smart Card V2 diagnostic interface plugged in.
If the ULA is the suspect pulling all the current as this is the only IC that is incredibly hot after 5 mins, then there's little chance
of obtaining a another. If it runs well with a current limit then I suppose that is a cheaper fix then a replacement ULA.

[1024MAK]

The ULA is supplied only by the +5V (from the 7805 voltage regulator), so it is definitely not the cause of your problems. It is designed to run hot. Internally the logic runs on a 0.85V to 0.95V supply so there are on-chip series voltage regulators to drop the supply voltage (pin 13 which should be around 3V to 4V) down. This is the cause of a lot of the heat...

As I said earlier, the most likely suspects are the 4116 DRAM, or a fault with the DC-DC circuitry. Can your bench PSU output a regulated +12V supply?

If yes, see if you can isolate the +12V supply to the 4116 DRAM. Then use your bench PSU to supply it. Note that R59 HAS to continue to be fed by the DC-DC converter +12V rail, as this, along with R58 and TR5 are the feedback control for the DC-DC converter. A fault in this part of the circuitry can cause the DC-DC converter to try to run beyond the normal power level, thus resulting in TR4 overheating.

Also note that with an external PSU suppling the +12V supply to the DRAM, the power up sequence should be:
Power up the 9V supply to the ZX Spectrum.
Power up the +12V supply to the 4116 DRAM.
Reset the Z80 (short out C27).

If you have two multimeters, it would be helpful if the +9V current and the +12V current could be monitored

In answer to your question about the ‘coil’, if it has overheated in the past, then it is possible that a short has developed either between the primary and secondary windings, or within the same winding. Resistance testing is not usually helpful due to the low resistance of the windings of a good ‘coil’ so it’s not easy to tell if there are shorts within a winding.

However, as the DC-DC converter is kinda working, it’s difficult to say for certain that this is the problem.

Where the ‘coil’ is found or suspected to be faulty, either buy a new one (but they are expensive) or carefully unwind the old enamelled wires, then wind on some new enamelled wires. The diameter of the wire has to be the same, and the number of turns and the direction of the winding is very important. It has to be the same as the old one. It is also a good idea to use some varnish to stop it from vibrating, but you may want to leave this until after you have it working.

Mark

[Ast A. Moore]

Strange thing is that the excessive current draw is only when first applying power, after powering on I can disable the current limit and the speccy only draws around 750mA ?

It does suggest the coil might be the culprit. That would (partially) explain the reactive nature of the excessive current draw. I’d cobble together a three-rail power supply and test the machine that way, bypassing the original power supply circuitry altogether. That would pinpoint the problem.

[rimmeruk]

Well, its a relief to know the ULA is out of the suspect list, I did pull the ULA and test on a known working board anyway.
Just wasn't sure it was pulling any power from the +12V or -5V supply.

I decided I'm going to pull the coil and replace with a known working one and take it from there. This is the only component that
I have overlooked as the DC-DC circuit is working to some extent. There doe snot seem to be a short circuit as such as if there were
then this short would be present throughout the whole time the circuit is powered. The coil is the only component that would react in such a way, as the current is alternating and building up the voltage in the switching circuit when powering up.

Next I will isolate the 12V to the DRAM chips as this means lifting all the 12V pins one by one. I'm not one for desoldering
all the DRAM chips as it's too much of a job if I can run the speccy with an LM317 current limit circuit.

IF the coil turns out fine, then I may even replace it with a modern DC-DC switching module and isolate the unreliable spectrum's
power circuit. There are plenty of DC-DC boost/buck convertors on fleebay nowadays and they available in small modules..

My paitence is slowing running out with this board and even though I have a good knowledge of electronic components and circuit repairs, this DC-DC circuit is one of the most annoying and over engineered designs that I have had to diagnose. I suppose it was a great design in the 80's but not very efficient or practical by today's standards and has much room for improvement

I'll keep you all posted.

[zx64]

The ULA is supplied only by the +5V (from the 7805 voltage regulator), so it is definitely not the cause of your problems. It is designed to run hot. Internally the logic runs on a 0.85V to 0.95V supply so there are on-chip series voltage regulators to drop the supply voltage (pin 13 which should be around 3V to 4V) down. This is the cause of a lot of the heat...

Really? 40 years old chip has 0.85V to 0.95V logic?

[Ast A. Moore]

The ULA is supplied only by the +5V (from the 7805 voltage regulator), so it is definitely not the cause of your problems. It is designed to run hot. Internally the logic runs on a 0.85V to 0.95V supply so there are on-chip series voltage regulators to drop the supply voltage (pin 13 which should be around 3V to 4V) down. This is the cause of a lot of the heat...

Really? 40 years old chip has 0.85V to 0.95V logic?

The ULA is a strange, hybrid beast. There are a gazillion voltage regulators inside of it:

In a fully utilized array, the largest power consumers would be the peripheral cells. Because of this, the power rails were routed along the inside edge of the peripheral cells, surrounding the central matrix square. . . . Not only did this provide power precisely where it was needed, but was the most efficient way to supply the matrix cells by guaranteeing that no cell was further than half the matrix width from the Vs supply. . . .

Peripheral cells operated from Vcc, whereas matrix cells operated from Vs which, in the case of CML arrays, was a regulated, noise-free supply of between 0.84 and 0.95V, generated by temperature-compensated series regulators located at the base of each peripheral cell. . . . Since there are many Vs regulator evenly distributed around the outside of the array, the total current available was much greater than a single regulator could provide. . . .

All arrays could be power form a single supply, where the internal Vcc and regulator supply Vreg were commoned. Alternatively, CML arrays permitted the use of two separate supply voltages for Vcc and Vreg, producing a lower power dissipation without compromising speed.

That’s an excerpt from Chapter 5 of Chris Smith’s book The ZX Spectrum ULA. How to Design a Microcomputer.

[rimmeruk]

Interesting posts about the ULA, but that thing runs hotter than my Core2 2.6Ghz cpu so there's an awful alot of wasted power going on in there for something that runs at 4Mhz

Back to my topic, I think I have found the problem with the constantly blowing TR4 transistor. I was using a ZTX450 which is only rated at 1A and of course it should have been a ZTX650 or ZTX651 rated at 2A I feel rather embarrased actually as I should know better. I still don't understand why it would blow when the vanilla spectrum only uses around 750mA, it is still within the operating range for a ZTX450.

I read somewhere that the ZTX450 was a suitable replacement for transisor TR4 as the wattage is the same. I guess the HFE and bias voltages make a difference also in the boosting circuit as I compared both on my transistor tester and they are not the same.

I did swap the coil out and that was working fine, actcually both coils had 3 ohms on both the windings. Anyway, the spectrum is up and running great now from a proper DC power adapter, no spieks in current on power on either. I still don't understand why a ZTX450 would cause an inital current spikes but still operate fine for hours.

[zx64]

Yes very interesting info about ULA, I agree.

My ULA doesn't get hot very much, only slightly warm.

[1024MAK]

ULA info, points to keep in mind...

The ULA uses bipolar transistor technology, whereas modern CPUs use modern ‘CMOS’ technology. Bipolar transistor technology is much more power hungry compared to CMOS technology.
The ULA technology of this period, requires higher internal switching currents for faster speed operation. Lower power ULA types were available for low speed applications.

The ULA is working well within the temperature and electrical power ratings of the DIL package used. Semiconductor junctions can work without problems at between 100°C and 125°C. So what may burn your finger may be normal for the chip. The biggest problem is that the 7805 voltage regulator heats up the inside of case, a case with not much airflow.

The ULA type used in later board versions were a different series, they do run considerably cooler compared to the first type that was used.

The ULA does not run at just 4MHz. The primary function of the ULA is the video system. That’s a high speed speed system. Hence the main oscillator is 14MHz... (although note that this is divided down internally to lower speeds).

Mark

[1024MAK]

Back to my topic, I think I have found the problem with the constantly blowing TR4 transistor. I was using a ZTX450 which is only rated at 1A and of course it should have been a ZTX650 or ZTX651 rated at 2A I feel rather embarrased actually as I should know better. I still don't understand why it would blow when the vanilla spectrum only uses around 750mA, it is still within the operating range for a ZTX450.

I read somewhere that the ZTX450 was a suitable replacement for transisor TR4 as the wattage is the same. I guess the HFE and bias voltages make a difference also in the boosting circuit as I compared both on my transistor tester and they are not the same.

When comparing transistors in this kind of application, there are many parts of the specification and characteristics that have to be looked at. One important point is that with DC-DC converters, the switching currents are very different to the input supply, or output currents. In fact, the switching currents can be considerably higher. This is also part of the reason that the electrolytic capacitors in this part of the ZX Spectrum are critical to normal operation.

Also when switching an inductor (coil, transformer), when the magnetic field collapses, high reverse voltages will be generated.

You also want a transistor that has a high gain at a high current, and which has very fast switch on and importantly, a very fast switch off speed. If the transistor is too slow, it will overheat. Hence selection of the correct transistor is critical. ZTX651 and ZTX653 transistors are still in production, and are the only recommended replacements for TR4 (Sinclair used the ZTX650 which is a lower voltage version of the ZTX651, and which is obsolete).

Mark

[rimmeruk]

Also when switching an inductor (coil, transformer), when the magnetic field collapses, high reverse voltages will be generated.

You also want a transistor that has a high gain at a high current, and which has very fast switch on and importantly, a very fast switch off speed. If the transistor is too slow, it will overheat. Hence selection of the correct transistor is critical. ZTX651 and ZTX653 transistors are still in production, and are the only recommended replacements for TR4 (Sinclair used the ZTX650 which is a lower voltage version of the ZTX651, and which is obsolete).

Mark

Yes, seems it is important that the transistor can handle the higher reverse current as > 1A is being generated when the boost circuit is starting because that was the only time the transistor was blowing.