Noisy Spectrum...

Post Reply
bola_dor
Posts: 319
Joined: Wed Oct 02, 2019 5:32 am

Noisy Spectrum...

Post by bola_dor »

Hi. I finally decided to fix my Spectrum audio output. It's mute but sounds good using an amplified speaker.
I identified the problem at the internal speaker. It measured as open circuit. I thought it was an easy fix. Exchanged the speaker with a 32ohm one (those worked fine with other speccy boards. A bit louder perhaps) . But I now hear a medium/high pitch noise. Changed the transistor and the diode, and tried another speaker.. noise persists.. mic output is still clear.
Revieed the track from ULA pin 28 and looks good.
Put an oscilloscope to ULA pin 28 and it is really noisy (mic output isn't)
20210913_121504.jpg
Then I measured pin40 GND, oscilloscope gnd probe is attached to the heat sink.
20210913_100615.jpg
I measured the RF modulator chasis against heat sink gnd and got a similar picture..
ULA is a -7 version. Board is a CZ Spectrum (clone) that is mostly an improved issue 6a. (No 12v ram and RF has a composite output as standard)
Ernesto
ZX80, TK83, CZ1000, CZ1500, TK85, CZ2000, CZ1000Plus, CZ1500Plus, CZ Spectrum, TK90X, CZ Spectrum Plus, ZX Spectrum, ZX Spectrum+, ZX Spectrum +2, Sinclair QL. And more to come :D
User avatar
Tiger
Posts: 38
Joined: Tue Sep 01, 2020 10:37 am

Re: Noisy Spectrum...

Post by Tiger »

Maybe the noise is a problem with the power supply - collector TR 7 (at the real Spectrum) is directly connected to the +9V Input. I would try to take an other power supply ...
bola_dor
Posts: 319
Joined: Wed Oct 02, 2019 5:32 am

Re: Noisy Spectrum...

Post by bola_dor »

Tiger wrote: Tue Sep 14, 2021 12:38 am Maybe the noise is a problem with the power supply - collector TR 7 (at the real Spectrum) is directly connected to the +9V Input. I would try to take an other power supply ...
Yes. The board is very similar to the Timex/Sinclair and that part is the same. Transistor is a different part but can be safely replaced with a BC337..
I think the diode should prevent that noise to reach the ULA. PSU is somewhat noisy but not that much and noise is different at both sides from the diode. Your advice made me review all traces and connections and Transistor was misplaced.. (so thank you!) now I fixed it and have the internal speaker working but noise at GND and ULA pin 28 persists..
Should I concern about that?
Display image is good. Some minor jailbars, I wonder if should I ad bigger capacitors at C28 and C75? (That helped before with jailbars at real Issue 6a boards more than the 12v ram decoupling published elsewhere..)
Ernesto
ZX80, TK83, CZ1000, CZ1500, TK85, CZ2000, CZ1000Plus, CZ1500Plus, CZ Spectrum, TK90X, CZ Spectrum Plus, ZX Spectrum, ZX Spectrum+, ZX Spectrum +2, Sinclair QL. And more to come :D
User avatar
1024MAK
Posts: 3778
Joined: Mon Sep 26, 2011 10:56 am
Location: Looking forward to summer in Somerset, UK...

Re: Noisy Spectrum...

Post by 1024MAK »

Digital computers are electrically very noisy. This is due to multiple transistors inside the various chips switching on or off at multiple different rates.

This is why there are so many (decoupling) capacitors used (modern rule of thumb is at least one per chip).

Because the switching on/off (especially the driver outputs of the chips) is fast and they have to drive the parasitic capacitance of the relevant address, control or data line, this causes a short and sudden, but relatively high current to be drawn from the power supply rail. Also with some digital chips that use two transistors per stage, for a very brief moment in time, both transistors are partially on, causing a brief short circuit! All this in turn briefly causes both the voltage on the power rail to drop slightly at the chip’s power pin AND the 0V/GND pin voltage on the chips 0V/GND pin to rise with respect to the voltage on the board power input or voltage regulator 0V/GND connection. This is because the current flowing in the supply and 0V/GND PCB tracks causes a small voltage due to the sight resistance of the conductors.

In an ideal PCB layout, either a four layer board would be used with one layer being only used for a ground plane, or a star earth arrangement would be used, where each chip (or other active circuit component) would have its own 0V/GND connection back to a single 0V/GND/earth point (normally where a voltage regulator is used, it’s 0V/GND terminal, otherwise the first capacitor after the power input connection point to the board).

However, because the Spectrum was designed to be low cost, and less was known about power distribution problems, neither of these techniques were used. This did bite Sinclair with a long winding 0V/GND and hence noisy 0V/GND on the QL boards… and an official modification resulted…

And yes, replacing selected existing ceramic decoupling capacitors with more modern multilayer larger values or adding additional modern multilayer ceramic decoupling capacitors in parallel may help. It’s best to experiment rather than me guess as to which ones… By this, I mean, set the meter up with clips so you don’t have to hold probes, then hold a new capacitor across the leads/pads of an existing decoupling capacitor and see if if makes any difference. Repeat until you are happy or the noise drops significantly.

The upgrading of the decoupling capacitors on the +12V supply is useful in computers where 4116 (or equivalent) DRAM is used, because their main (highest current) supply is the +12V supply. And Sinclair used ONLY FOUR 22nF decoupling capacitors, which is rather stingy. In computers that use DRAM that only needs a single +5V supply, obviously only this supply needs decoupling capacitors.

Modern practice is to use 100nF multilayer ceramic capacitors, at least one per chip. In a ZX Spectrum, for eight 4116 DRAM chips, that would be 800nF compared to Sinclair’s 88nF!

The other important thing to know about decoupling capacitors, is that the leads should be as short as possible and they should be connected as close as possible to the power and 0V/GND pins of the chip.

Incidentally, 100 to 150mV of noise on the power rails/0V/ground is normal for a digital computer. A typical ZX Spectrum is not likely to be as good as this though.

Mark
bola_dor
Posts: 319
Joined: Wed Oct 02, 2019 5:32 am

Re: Noisy Spectrum...

Post by bola_dor »

Thank you Mark,
As you ca see there is at least one ceramic per IC although these are mostly 22nF each.
20210914_191313.jpg
Aded some bigger ones to ULA vcc/gnd (100uF electrolytic and 1uF mlcc) didn't help much as with other boards..
As you can see layout is very similar to Issue 6a.
Ernesto
ZX80, TK83, CZ1000, CZ1500, TK85, CZ2000, CZ1000Plus, CZ1500Plus, CZ Spectrum, TK90X, CZ Spectrum Plus, ZX Spectrum, ZX Spectrum+, ZX Spectrum +2, Sinclair QL. And more to come :D
User avatar
1024MAK
Posts: 3778
Joined: Mon Sep 26, 2011 10:56 am
Location: Looking forward to summer in Somerset, UK...

Re: Noisy Spectrum...

Post by 1024MAK »

bola_dor wrote: Tue Sep 14, 2021 3:57 am I wonder if should I ad bigger capacitors at C28 and C75?
Keep in mind the ULA has two supply inputs. Pin 13 is the supply for all the matrix logic cells. Pin 14 is the 5V supply for the interface/peripheral cells that connect between the matrix logic cells and the pins.

Also, what type of PSU are you using? An old original type? Is it an unregulated type?

Or are you using a modern switch mode power supply unit (SMPSU)?

Have you tested the ripple and noise on the +9V (nominal) supply? What about on the +5V supply? At at ULA pins 13 and 14?

I won’t have time today, but later in the week, or at the weekend, if you think it will help, I can post up some images from my ‘scope of a Sinclair Issue 6A board.

Oh, also, check/test/replace C35 (10nF) and continuity test from its legs to 0V/GND and to ULA pin 28.

If you want to hear a real noisy computer speaker, get an Acorn BBC Micro. In amongst the (surprisingly loud) general hiss and noise, it also produces noises and sounds that sound like some special effects from a budget sci-fi movie… :lol:

Mark
Lardo Boffin
Posts: 1509
Joined: Sat Nov 26, 2016 2:42 am

Re: Noisy Spectrum...

Post by Lardo Boffin »

And the Oric 1 has quite a noisy speaker!
ZX80
ZX81 iss 1 (bugged ROM, kludge fix, normal, rebuilt)
Iss 3 ZXVid
TS 1000 iss 3, ZXPand AY, ZX8-CCB, ZX-KDLX & ChromaSCART
TS 1500 & 2000
Spectrum 16k (iss 1 s/n 862)
Spectrum 48ks plus a DIVMMC future and SPECTRA
Toastrack
Post Reply