Mixerton HiFi Amplifier Balanced Input Addition

David J Greaves. University of Cambridge, March 2020.

Under construction ...


Final grand view.

In this blog I report on changing a domestic HiFi amplifier to connect to professional-audio, balanced XLR connections. Lock down had prevented access to my normal set up.

If you have connected your PC to a domestic HiFi amplifier, you might experience an annoying background sound that varies according to what is running on the computer. It might even `squeak' as the mouse is moved. This can happen both within internal sound cards and with external audio interfaces. In most cases, the primary cause is earth currents. These are avoided if nothing that is earthed is also connected to the amplifier and the HiFi amplifier itself is not earthed (they are normally double-insulated instead of being earthed). Laptops fair better, since even when plugged in to the mains adaptor, nearly all are not earthed through that (even if they have the 3-pin `mickey mouse' IEC C5 clover connector). See companion article: Studio Wiring and Earth Loops.

Unlike many consumer HiFi separates, the amplifier I used here was grounded (mains earth connected/bonded to signal ground).

Professional-quality sound cards, whether internal or external, tend to have balanced outputs. These are much better at overcoming earth loop noise than the single-ended outputs found on the 3.5mm mini-jacks used for PC audio on laptops and motherboards. But the balanced advantage is lost when the balanced signal is simply connected to a domestic HiFi amplifier with its typical single-ended RCA phono sockets. The simple connection will directly connect the 'cold' signal to the amplifier ground which unbalances the system, largely preventing cancellation through signal subtraction from taking place.

Original Amplifier Structure

Original Back Panel
Original Back Panel.

The amplifier I modified for this article was an old Cambridge Audio A1 mark 3. This is an earthed amplifier, with signal ground connected to mains protective earth from the middle of the power amplifier and also, via a small resistor/capacitor pair, at the RCA socket point. Otherwise, this model has a very straightforward design that was perfectly engineered for its target market and price point. The original back panel had a large number of phono sockets. I was only using one input pair. I was also using the tape output pair to feed a larger, professional amplifier. I was getting a most annoying level of computer noise, especially when the output level from the sound card was on medium settings and the volume control on the amplifier was at higher levels. It was not so bad when the sound card level was higher and the amplifier volume turned down, which gives the same overall volume but uses a larger signal to earth noise ratio between the units.

Original input selector card
Original input selector card.

The amplifier had a large amount of spare space inside it. A large input card carried all the signals to and from the RCA sockets to the input selector and tape monitor switch. The card also has a site for a phono amplifier to be installed, but that was not fitted on my unit (whole amplifier was £48 on E-bay).

Rear View after Modification
Rear View after Modification.

I removed the back of the input selector card with a hacksaw. I drilled out the back panel and installed two XLR input sockets and two XLR output sockets. I used a dual op-amp on Veroboard to make a differential XLR to single-ended convertor. Its output feeds to the existing pre-amplifier input. The input selector switch was left completely disconnected. The tape monitor switch was re-used to serve as an enable/disable for the XLR output sockets.

Gain Scaling

The main power amplifier uses a TDA1514A.pdf. This was set to closed loop gain of 20680/680=30=30dB by 680R and 20Kohm feedback divider resistors. (This is the one gain figure that is the same in voltage ratio terms as it is in decibels!)

The pre-amplifier uses two NE5532 dual op-amp devices as a Baxandall tone control pair and a 6dB gain stage. The power amplifier rails are approximately +/- 28V. The maximum output amplitude before power amplifier clipping, when unloaded, was also about 28V, demonstrating close to rail-to-rail swing for the TDA1514. (Rather than straightforward flat clipping it produces quite a strange waveform when overloaded owing, perhaps, to its bootstrap and protection circuits. At least it was at equal positive and negative levels.)

The input level was being changed from -10dBV to +4dBu which is a difference of 11.8 dB. The system gain therefore needs reducing by the corresponding amount, assuming it was roughly correct before.

The new input stage has a gain of 6k8/7k8 which is approximately 0 dB.

To change the overall gain, I took out the 20K feedback resistors in the power amplifier stage (pin 5 to pin 9 of TDA1514A) and replaced them with 4k7. This gives closed-loop gain of x7.9 which is approximately one quarter of the original x30, as required.

The overall gain is thus about 16.

Circuit Details

One channel of the differential input and power supply regulators
One channel of the differential input and power supply regulators.


The circuit is basically just the standard differential op-amp circuit. I used a TL072 since I had one to hand and the better noise levels of an NE5532 do not matter at +10dBu.

Input capacitors are used to suppress RF pick up on the cable runs. Such signals can rectify in input transistor non-linearities to produce in-band noise and related problems. But the input capacitors are a major potential source of channel imbalance. This can ruin common mode rejection. Ceramic capacitors are excellent at handling RF pickup, but they do have poor tolerances (20 percent or worse). The 22pF capacitors do not have any in-band effect, so their tolerance does not matter much. The 470pF capacitor starts to have an effect in the audio-band, but because it is across the pair, it effects both hot and cold equally and does not upset balance.

The output sockets are taken from the input socket signal with a series 1Kohm resistor in both the hot and the cold signal. This is not an ideal circuit configuration and could lead to some in-band HF roll off depending on the cable capacitance and RF suppression at the next piece of equipment. I put these resistors in place to provide the output mute function. The tape monitor switch was only two pole, whereas four poles would be needed to cleanly disconnect the stereo output. Rather than changing the switch or adding a relay, I just used the existing poles to short the output. This explains the 1Kohm resistors: it would be no good to also short the input.

The power supply rails are at too high a voltage to run a TL072 and it is also good to have a fair degree of isolation between the power-amp and pre-amp stages. The pre-existing pre-amp would typically have a suitable power supply, but I found on this design there are separate pairs of regulators for each of the two dual op-amp ICs. Which to choose? Also, these were nicely tucked under their little tin cans and the supply already routed to the input selector card was the raw power rails. So I added my own pair of 78/79L15 devices. I also put a couple of series 180R resistors in place to avoid massive current flows in case of a wiring mistakes. These provide additional isolation of course, which is also great. As it turns out, there was a wiring mistake: countless pin-out images of the 79L15 on Google images search are wrong and I happened to refer to one of the wrong ones when I was soldering up the board.

Top view of new input ampifier, constructed on Veroboard.
Top view of new input ampifier, constructed on Veroboard.


I made some brief measurements of the modified amplifier.

The amplifier gain, with volume control at maximum and tone controls flat was x14 as per calculated above.

The crosstalk between channels under the same settings was 69 dB. Why people ever measure or quote this for a stereo pair always seems a bit odd to me, since a figure above 25 dB is achieved by even the worst design that is not actually faulty and is surely quite sufficient!

The common mode rejection for the balanced inputs was measured as 42dB for the right channel and 56 dB for the left channel. With a resistor tolerance of 1 percent, 40 dB is all that can be expected. I was slightly tempted to trim the design to get above 50 dB on both channels, but I knew 42 dB would solve my original problem.


As noted above, unlike most consumer HiFi separates, this unit is grounded. This was causing an earth loop problem. See separate note: Studio Wiring and Earth Loops.


The modified amplifier works perfectly. I had nearly all the parts to hand, so the cost was a few hours of work. This provided a welcome alternative to doing my real work during lock-down. I no longer hear the computer noise. Compared with professional audio monitors, monitoring on HiFi can provide an insightful tonal alternative. Monitoring in mono is also a great feature, so perhaps I will next add a mono switch? I've not traced out the circuit on this unit, but in most amplifiers, a mono switch can easily be added to the balance control circuit.

See also build-your-own-professionalgrade-audio-amp-on-the-sort-of-cheap.

This material is provided in good faith, but you use it at your own risk. I will deny any liability for any losses or infringements that may arise.