Bush DAC90a Modifications - Sound Quality and Scale Lamps

[kalee20]

Earlier this year I won on eBay, a Bush DAC90a, described as ‘working and original condition’. On arrival, and believing it had recently been used, I felt confident enough to plug in and switch on – it did indeed work, and I was very pleased with sensitivity and selectivity, but not so much with sound quality which was really muffled. Feeling sure that there was room for improvement, I spent a fair time analysing and testing the circuit, and came up with these modifications. I’d like to think that the only reason that I was able to improve on the Bush designers’ work was that I was freed from the constraints of designing to a budget on a tight timescale! As a matter of detail, the radio packed up with a fizz and a phut as the across-the-line RFI capacitor failed, within an hour of use – and apart from some chips in the knobs and a replacement of ‘THAT’ capacitor, the description of ‘original condition’ turned out to be accurate! And all the waxed paper capacitors were leaky-ish, and all carbon resistors read 20% high!

SCALE LAMPS

In use, the scale lighting was marginally better than having a couple of cigarette butts glowing behind the scale. The lamps fitted were correct types (3.5V @ 0.15A), paralleled with 75 ohms.

Looking through DAC90a threads on this forum, it appears that when first launched, there was a parallel resistor of 250 ohms, but within a year or two this was lowered to 75 ohms, and at the same time, the rectifier limiting resistor changed from 150 ohms to 250 ohms. Why? I can only surmise that although the original values were OK, perhaps there were many bulb failures caused by people switching off and then back on a few seconds later, and the HT charging surge (which flows through the lamps) burned them out. So, as a compromise, I raised the 75 ohms lamp shunt to 180 ohms, and replaced the UY41 limiting resistor with its original value of 150 ohms. This gives a current slightly higher (in RMS terms) through the lamps, and slightly less heat generation in this limiter. Although the UY41 requires 150 ohms minimum, I reasoned that the DAC90a demands only about half of what the valve can deliver, and moreover there is additional limiting resistance in both the heater dropper (on 230 and 250V taps) and by the lamps themselves. So, I wasn’t worried about running the valve at its limits.

Finally, I replaced the translucent plastic diffuser (which was yellowing and wavy, and made from inflammable cellulose nitrate, as a match test proved ) with a piece of 0.5mm white translucent polyester sheet (PM me if you would like some!). With these mods, although not bright, the scale can now be seen in daylight yet the rare bulbs are still comfortably under-run!

SOUND QUALITY

A quick sweep with an AF signal generator, from volume control to loudspeaker, revealed a treble cut with 3db corner frequency of 1.6kHz. No wonder the sound was so lacking in treble! Looking at the circuit, there is a 0.003µF capacitor from AF amplifier UBC41 anode to chassis, which seemed far too big.

Using the nominal value (42k) for the ra of the UBC41, the treble cut-off frequency of this stage calculates at 1.62kHz thus vindicating the measurements. As even with the present-day savage treble cut on AM broadcasts, there is still decent top-end to 4kHz, and this seems just plain the wrong value. As at some point I want to use a local micropower transmitter with decent treble, I calculated the value of capacitor to give a 3db cut-off of 12.5kHz, and arrived at 390pF. I replaced the 0.003µF capacitor with this value, fitting it directly on the UBC41 valveholder (as it also forms a bypass for any residual IF).

Next, I looked at the UL41 output stage. This has the conventional tone-corrector capacitor across the output transformer, to compensate for the rising impedance of the loudspeaker at high frequencies. But, it ultimately gives a treble roll-off. How much better, I thought, if I can arrange the load presented to the UL41 to be constant and resistive, using a series RC pair forming a Zobel network?

Snipping out the 0.01µF tone corrector, I applied the AF signal generator to the UL41 grid, and measured the AC voltage across the output transformer primary. Reasoning that the UL41 could be regarded as a constant-current generator – especially with the unbypassed cathode resistor which is a feature of the DAC90a – I connected a decade resistor box across the output transformer primary. At 200Hz (low enough for the speaker impedance to be regarded as resistive), I found that the amplitude dropped to half when the decade box was set to 4.7k. Therefore, the reflected speaker impedance must itself be 4.7k. Next, I disconnected the decade box, swept the frequency, and plotted the response characteristic. Lo and behold, this showed a break point of 4.5kHz above which the response rose with frequency. Therefore, the load presented to the UL41 must be equivalent to 4.7k in series with 0.166H – and so paralleling this with a series 4.7k and 0.0075µF should give a constant load (4.7k) at all frequencies. I used 0.0068µF and 4.7k as nearest standard values.

So much for treble response, but what about bass? The above paragraph ignores the rising impedance at low frequencies due to loudspeaker bass resonant frequency – and with the negative current feedback of the unbypassed cathode resistor, although distortion may be low, there is almost no damping of the speaker’s resonance. Not good. Replacing the current feedback (adding a cathode bypass capacitor) by voltage feedback from the output transformer secondary would fix this, except that DAC90a’s left Bush with a completely floating secondary/loudspeaker for safety. To connect the transformer secondary (and hence the speaker) to the live chassis would definitely be a no-no. I considered winding a custom transformer with a tertiary winding purely for feedback, and dismissed it as unworthy. There remained the prospect of using a bit of voltage feedback direct from the UL41 anode, but then the output hum would increase as the transformer would be operating between the rippley HT rail and the now-stiffened-up UL41 anode. Not easy! Increasing the HT reservoir capacitor would reduce mains ripple but give the UY41 a harder time; adding extra LC smoothing was out as I did not want to start bolting on extra LF chokes and capacitors; shunt-feeding the output transformer would do the job but would need the same order of size choke and coupling capacitor. Yet, a lash-up did show the prospect of better sound quality with voltage feedback.

The solution I adopted, as there is plenty of reserve gain, was to leave the current feedback in place (the cathode resistor unbypassed), add a bit of voltage feedback (a gain reduction of a couple of db), and inject into the cathode a fraction of HT ripple voltage. In theory this can neutralise the hum completely. I found empirically that 7k from the HT rail (via a DC blocking capacitor) gave a hum null, and therefore used 7.5k as the nearest standard value.

UL41 LEAKAGE

Although the grid coupling capacitor had been replaced, my UL41 showed a reading of about 0.4V positive on its grid. As it happens, the anti-parasitic stopper resistor is connected between the grid tag and another, unused, nominally spare tag on the UL41 valveholder. So, any internal leakage which develops within the UL41 button-base has its effects augmented. Configuring the layout to avoid connecting grid-related circuitry to sparee pins resulted in a drop to 0.15V positive on my UL41 grid – which I was contant with. Incidentally, during my lash-ups and ‘scoping around, I found that the anti-parasitic stopper can’t be dispensed with. However, I was able to reduce its value from 47k to 4.7k (which is good news as the voltage-feedback network increases the grid-chassis resistance by 120k)

AVC

The values of AVC bypass capacitors seemed a bit low to me (0.05µF). Although this yields a low frequency cut-off of 3.2Hz with a series AVC resistor of 1M, I increased the decoupling capacitors to 0.1µF as any bass AF on the AVC line would give hard-to-predict (but probably non-linear) effects, and the original capacitors were leaky anyway.

UBC41 CATHODE BIAS

The AF amplifier triode, UBC41, has cathode bias which also gives AVC delay. Fine – except that for some reason Bush used a resistor shared with the UF41 IF amplifier. So, when a strong station is tuned in, the AVC voltage rises, the UF41 is biased back, its cathode current falls, and the bias to the AF amplifier drops! Definitely not what is wanted. So, I separated the two cathodes, leaving the UF41 with its original value cathode resistor, and the giving the UBC41 its own cathode resistor. As the UBC41 cathode current is quite small (only 160µA), I found that 6.8k was appropriate to give the same bias voltage as before (under no-signal condtitions). It’s bypassed with 220µF.

FINALLY

All the above modifications were incorporated, the new circuit being as per the attached pdf file with changes in red (for comparison, the original is available on Paul Stenning’s excellent DVD-ROM). The extra components (3 capacitors, 5 resistors) easily fit using modern-day parts, as the under-side chassis photo shows. I did drill 2 holes and fit a small tag strip at the underside of the chassis. The quality is much better and I am really pleased. The cabinet has been cleaned and polished; the knobs repaired; the variable capacitor anti-vibration grommets replaced; the loudspeaker cone suspension centred and reglued; the rubber wiring replaced with silicone; etc. For the purists, I’d like to think that Bush would have done as I have, if they had been allowed. For future custodians, I’m leaving a radio with modern, highly reliable components, which should need no attention for many years, and there’s a copy of the new circuit tucked inside in any case!

[matthewhouse]

Wow thats really neat, I like the way you have sleeved all the bare wires. Some very useful information there.

[stevekendal]

Wonderful piece of research and immaculate wiring there, makes my own fumblings look rather crude I've done a few DAC's myself and they nearly always sound better with the cap across the o/p primary in the workshop bin and a smaller one from triode to chassis. Regarding the filter cap that goes bang, I think Gerry Wells cuts it out and leaves it out. Yes, the speaker suspension often comes unglued and I dont think you can get new ones any more that will fit these sets. Many thanks for sharing this with us. Steve.

[Ed_Dinning]

Congratulations Peter, this must be the definitive guide to DAC 90 mods.

Best regards, Ed

[kalee20]

Hi thanks folks!

Regarding 'sleeving all the bare wires': I used Vidaflex sleeving (woven glass-fibre impregnated with coloured silicone rubber). Looks the part 100%, but doesn't go sticky, immune to soldring-iron temperatures, etc. I also sleeved the yellow polyester capacitors with clear heatshrink: they look better (to me), you can still read the values, and there's no chance of the outer tape ever unwrapping!

Regarding the RFI filter capacitor: I replaced this with an X rated part, held down with a little brass strip, but didn't feature it in the mods because it's not a change in value. I have found occasionally I get a reduction in mains-borne interference in AM radios by adding an across-the line capacitor - it depends on the local installation!

Incidentally, are there any observations on my thoughts above on why Bush changed the dial-lamp bypass and UY41 limiter resistors?

[Alf]

Would I be correct in assuming that these modifications would also be suitable for the DAC10 as the circuit is identical, except for the obvious push buttons.

[igranic]

Congratulations. An excellent, meticulously engineered piece of work.


At a more trivial level:

Originally Posted by kalee20

I used Vidaflex sleeving (woven glass-fibre impregnated with coloured silicone rubber). Looks the part 100%, but doesn't go sticky, immune to soldring-iron temperatures, etc.

Where do you buy this sleeving? I have not been able to find it in the catalogues of the usual suspects (RS, Farnell, CPC).

[kalee20]

Hi Igranic,

I am fairly certain that the Vidaflex comes from Jones Stroud Insulations (www.jsins.co.uk), or from insulated Sleevings (NW) Ltd. (01253790911).

They will probably impose minimum order quantities.

I have also seen similar sleeving on a Suflex-labelled reel. Try a google search! I will have a word with our Purchasing people at work when we return, to find out for you.

[Boom]

Absolutely magical work kalee. Well done indeed. I was toying with the idea of doing similar and chrome plating the chassis and loudspeaker while it's all in bits.

[kalee20]

Originally Posted by Alf

Would I be correct in assuming that these modifications would also be suitable for the DAC10 as the circuit is identical, except for the obvious push buttons.

Alf, I'm not familiar with the DAC10 having only seen one and heard none!

However, looking at the DAC10 circuit diagram, I'd say yes, generally. (The DAC10 does have an unusual loudspeaker, which may colour the frequency response).

I'd start by reducing the UBC41 anode-chassis capacitor, which is 0.004µF in this set, and secondly, disconnect any UL41-grid related components from "unused" anchor tags on the valveholder. After that, implement other mods as you wish!

[razor600]

Excellent job,looks very neat.
Ray....

[PJL]

Bush must have had a reason for applying so much HF cut. All I can think of is electrical interference? Excellent job Kalee and really good to see the reasoning behind it.

My only question relates to the modification to the tone correction capacitor. I always thought that this also served to prevent transients that cutoff the output valve from breaking down the output transformer insulation?

[kalee20]

Thanks for comments, PGL.

I am honestly wondering, was this a typo by Bush designers which never got picked up and corrected? Certainly, there seems to be no justification for a triode anode bypass capacitor as large as 0.003µF. But as the radio does work with this value (albeit a bit muffled), it could easily have been overlooked right through production life.

Reference the tone correction capacitor, that's a good point. However, even with the resistor in series, if the output valve is abruptly cut-off, the transient voltage on the anode will be limited, just to not quite such a low value as with only the capacitor in circuit. If I have a transformer break down, I'll report this on the Forum - for me, it's not a big disaster as I do have access to winding facilities.