This latest build is the PeeCeeBee, a discrete Class AB amplifier, now in its fourth iteration. It was designed and made available by Shaan through a group buy over at DIYAudio. He has recently designed a very nice PSU as well and so I ordered a couple, one of which powers this amp. Although Shaan supplies the boards at a very good price, the shipping from India is expensive. The amp is described as a lateral MOSFET based Class-AB amplifier with symmetrical current feedback. I don’t pretend to understand entirely what a current feedback amplifier is, even after reading about it, but this topology is well-regarded in some circles.
Specs, as provided by Shaan, look impressive (measured with generic components/passives and basic unregulated linear PSU):
Maximum Power – 50WRMS/90WRMS into 8Ω/4Ω load with +/-35V PSU
Maximum allowed PSU voltage for 4Ω operation – +/-35V (with default clamp diodes and BD BJTs for Q9/Q10) and +/-42V (with higher voltage rated clamp diodes and MJE BJTs for Q9/Q10)
Frequency response (simulated) – ~2Hz to ~800KHz (-3dB)
AC gain – 23 (27dB)
DC gain – Unity
Input level for 50WRMS into 8Ω – 1VRMS
THD 100Hz 10W – 0.0005%, 1KHz 10W – 0.001%, 10KHz 10W – 0.002%
Slew Rate – 100V/uS
SNR – >100dB
Offset variation – +/-10mV
For this project I wanted to make it as good as I could within the limits of my funds. It has to be sonically and aesthetically pleasing and I didn’t want to cut corners unnecessarily. I didn’t have the funds to complete this build in one go, so it has taken a few months to get to this stage. I ordered all of the board components earlier in the year and the chassis, transformer, extra circuitry and panel hardware came in stages later.
The boards are beautifully designed and thicker than normal, constructed of 2.4mm FR4.
The fully populated boards are shown below. Note the transistors soldered to the underside for contact with the heat sink.
The pictures below show the PSU populated, except for the rectifier diodes, the spaces for which you can see at the edge of the board in the first image. I had trouble finding the vertical ‘box type’ resistors located near the relay, so I used wire-wound instead and bent the leads in to fit.
Component selection for the PSU is largely straightforward if you stick with the standard values of 35V rails into 8Ω, but there are some things to be aware of. The PSU board incorporates independent two channel discrete speaker protection circuits powered from the rails, so the values for resistors R9, R10, R11, R12 have to be calculated depending on your rail voltage and relay specifications. I chose the Omron G2R 1 E Relay (24V, 260Ω) just because it looks so good!
Care should also be taken with the transformer values. Mains supply voltage can fluctuate by up to 10% (in the UK at least), so although I bought a ±25V 300VA toroidal, aiming for 35V rails (25V X 1.4), the voltage at the rails can be as high as 39V (27.5 X 1.4). The calculation is simple and is provided in the schematic, but I have provided my steps below, which take into account the possible higher voltage:
R9, R10, R11, R12 values =
[ (Vrail-Vrelay) / (Vrelay/Rrelay)] /2
Vrail = PSU rail voltage; Vrelay= Relay coil voltage rating; Rrelay= Relay coil resistance
[ (39-24) / (24/260) ] /2
[15/0.09] / 2
166.6 / 2 = 83.3Ω
You just need to choose the next highest standard value, which in my case was 100Ω 1W.
The picture below was taken later and shows the rectifier diodes. There is a slight problem with the design of the PSU board layout in that the diode placement is too close together and they will touch each other if placed upright (this has been rectified in a later board revision). To account for this, I placed the diodes under the board, allowing me to fan them out more easily. They can also contact the chassis base plate for cooling purposes.
The first step in construction is always to check the power supply. For this I also included the Hypex softstart. The result was 38V, so my earlier concern about fluctuating voltage and choosing to factor this in when calculating resistor values was justified.
The next stage is to mount the amplifier boards onto the heat sinks. Shaan provides a drill template, but I had to modify the drill holes slightly due to alignment of the transistors. I will be using M3 x 0.5 machine screws, requiring 2.5mm drill holes for the M3 tap. I hadn’t used a tap since I was at school over 30 years ago, so there was a feeling of trepidation before I started, but it went smoothly.
Once the amp is connected to the heat sink, the setting up commences. This comprises four stages;
1. Setup and Check
2. VAS (Voltage Amplifier Stage) biasing
3. Offset trimming
4. Mosfet biasing
As I understand it, biasing is the process by which the optimal voltage or current is set which allows the transistor to operate within certain parameters and is crucial for low distortion.1 The offset trimming corrects for DC offset, which is the deviation of the symmetrical AC waveform (the positive and negative peaks are not equidistant from zero). A perfect measurement would be zero mV.
Stage 1: The tim pots VR1 & VR2 are set to maximum resistance and VR3 to zero resistance and the jumpers J1 & J2 are kept open. A 1Ω 2W resistor is placed in series with each power rail to the amplifier. Multimeter leads are placed across one of the 47Ω resistors. A brief power up to check if the resistors get warm or fuses blow and so on and a check of the MM reading.
Stage 2: The VAS bias current needs to be set at 10mA which results in a voltage reading of 470mV. VR1 & VR2 are adjusted to achieve 450mV, which will rise once warm.
Stage 3: Probes are moved to the output and GND terminals and again the two trim pots are adjusted to achieve an offset within +/-2mV of 0V.
Stage 4: Probes are moved to the 1Ω rail resistors and VR3 is adjusted to achieve 100mA Mosfet bias which is equal to 115mV. As the amplifier warms the reading rises, so a reading is taken after 15 minutes or so.
Once the adjustments are made the jumpers are closed, the rail resistors are removed and a test speaker is attached, as seen in the video below.
I had a problem with the second channel, a product of my soldering. I had shorted one of the BD transistors which led to a resistor smoking. Advice from Shaan was that a small transistor (Q3, Q4) was most likely damaged. I checked for damaged BJTs by measuring the base-collector and base-emitter with the diode test function of the multimeter. You need the datasheets to see the pinouts for each type of transistor and the voltage saturation values. For the NPN devices the red lead connects to the base. You then connect the black lead to the collector and emitter in turn. The values displayed are the forward volt drop in millivolts across the junction when probing the collector and emitter. These values should match the saturation voltage from the datasheet. For the PNP transistor check, the black lead is connected to the base.
After replacing the dead parts, the amp powered up nicely and I was able to get it biased without further problems. I was surprised I’d missed the short as I usually check every joint after I’ve soldered it, but I take it as a warning against complacency.
I find the layout and wiring of the components is the most difficult aspect of construction. For this build I wanted to pay particular attention to the grounding to reduce hum and noise. I overlooked this on my LM3886 as I had noticed no discernible noise at first. However, this was likely due to only connecting battery-powered sources, such as phones, tablets and a mains powered Chromecast. Once my daughter connected her laptop to it, the noise was terrible when it was plugged into the mains. I wanted to reduce the possibility of this happening from the outset and so fell into the confusing and often contradictory world of grounding.
Two resources I found were particularly helpful in understanding this aspect of DIY Audio. The first is an article on the website Elliott Sound Project by the ever helpful Rod Elliott: Earthing Your Hi-Fi – Tricks and Techniques. From here I got the idea to include a ground loop protection circuit. The rationale behind this is that the loop breaker adds a resistance in the earth return circuit, which reduces any circulating loop currents to a very small value, and ‘breaks’ the loop. The capacitor helps prevent radio frequency interference and the diode bridge provides the path for fault currents. For more detail on the various causes of hum, buzz and noise in amplifiers and their relationship to grounding, with considerable practical advice, this post and associated material from the HiFiSonix website is invaluable.
The build of a LM3886 chipamp on the website circuitbasics.com has used the Rod Elliott protection circuit and provides a link to a PCB layout that you can edit and order online for fabrication. I ordered it with no modifications and 10 PCBs came to £6.00 including postage from Hong Kong. Images of the bare and populated board are shown below.
A short note on extra circuitry. As mentioned earlier, I didn’t want to cut corners only to have rectify shortcomings later. A loudspeaker protection circuit (a circuit that interrupts potentially dangerous DC flow to the speakers) is provided on the PSU board, but I also included a soft-start circuit. These circuits are also known as inrush current suppression, and limit the initial current drawn when a power amplifier is switched on, thus protecting various components. Detailed information is available from ESP. For this amplifier, I chose the Hypex module as it is well-regarded but also provides the circuitry that allows the use of an LED momentary power switch. The modules are fully populated and come with comprehensive usage instructions.
The layout of the wiring and components is detailed in the block diagram below. Note that only connections between the components are shown, not the internal circuitry. As I am still very much a novice, I drew a sketch of the layout on paper and then drew it digitally in Adobe Illustrator. The reason for this is so that I have nice, clear editable drawing and I can show it to people for feedback. The first iteration had the signal input and return connected to the amp and the ground from the PSU was separate to the safety ground. I posted this version to DIYAudio forum for feedback and after a little back and forth with help from forum members including the amp designer, Shaan, I had the version you see below. Here, the power ground and signal grounds are jumpered together before connecting to the PSU ground. To help reduce noise, a screened cable serves to transfer the audio signal to the amps. However, the RCA input ground tabs are connected together and these are then directed to the PSU ground rather than the –0V input on the amplifier. You can see in the drawing that the RCA inputs are separated by the AC socket. This is for convenience in the drawing – in reality the finished amplifier has these inputs close together to reduce hum.
The PSU is acting as a star ground as I understand it, with just one connection leaving this point towards the ground loop protection circuit, before returning to the chassis at the safety earth. It took me quite a while to get rid of a low-level hum, but a simple re-routing of the left channel signal input away from the AC input and soft start solved it, even though the cable is now longer and right next to the transformer!
On my last two amplifier builds I used a power switch integrated into the IEC inlet, and this was placed at the rear of the chassis. I like this approach for its minimalist aesthetic, but it is awkward to use. For this build I decided to use a front mounted switch and chose one of the ring LED vandal proof designs. This comes with five connections;-the three centre pins are marked as NO (normally open), NC (normally closed) and C (common). The outer pins are for the LED and unhelpfully, these were not marked. A check with the multimeter set to Diode mode applies just enough voltage to illuminate the LED when you get the correct polarity. Unfortunately, as far as I can tell, the wiring to the soft start module has the LED always on, a de-facto standby mode regardless of the power state. There is a way to wire two different colour LEDs to indicate power on and standby states, but no way that I can see for the LED to go off without unplugging or having another mains switch.
The enclosure comes from Modushop. This time I used the Dissipante 3U 300mm with a 10mm front panel and 3mm aluminium covers and rear panel with the holes for the inputs and power switch drilled by myself.
As for the sound, I have it connected to a Schiit Modi Uber DAC and Magni 3 headphone amp which I’m using as a preamp for the moment (I’m planning on building a Rod Elliott ESP preamp next to go with the PeeCeeBee). I think the sound is just superb. I swapped between the PeeCeeBee and a Cambridge Audio amplifier that I’ve been using previously and for me there are two noticeable differences. I am useless at describing sound, but the bass certainly feels tighter and punchier, it seems to have less of a bloom than the Cambridge. The mids seem to have a little more clarity, almost sharp, but not in a harsh way, but this is less obvious than the difference in the bass.
This certainly wasn’t a cheap build, costing in the region of £400, but I am very happy with it.