LM4780 Audio Power Amplifier - Glenn's Design Log

[ Back to LM4780 Amplifier Page ]   [ Glenn's Audio and Hi-Fi Page ]

  • 12 April 2004
  • In the beginning
  • 4 May 2004
  • Sample chips arrive
  • 14 May 2004
  • Test power supply
  • 27 May 2004
  • Extended primary on power supply transformer
  • 1 June 2004
  • Prototype constructed on perforated board
  • 28 June 2004
  • "NC" pins are not connected
  • 26 August 2004
  • Role of the "GND" pins
  • 27 August 2004
  • LM3886 equivalent schematic details
  • 24 September 2004
  • LM4780 is a dual LM3886
  • 17 January 2005
  • Prototype demonstrated to Melbourne Audio Club

    12 April 2004: In the beginning

    I first became aware of the LM4780 chip in a brief product announcement in Australian Electronics News. The Australian distributor is Arrow Electronics, who have sales offices around the country.

    4 May 2004: Sample chips arrive

    The two sample chips which I ordered from National Semiconductor arrived from their office in Hong Kong.

    Photograph Copyright Glenn Baddeley 2004
    A back and front view of a pair of LM4780 chips. They are sitting on a photographic scale which has a row of 10 millimetre black and white squares. The 27 pins emerge from the chip with a 1 millimetre spacing. They are pre-bent into a 14 pin front row and a 13 pin back row which are 4 millimetres apart. The pin spacing in each row is 2 millimetres, which at 0.0788" is slightly smaller than the 0.1" pin spacing used on most chips. An Australian $2 coin is included in the foreground for scale for the Australian folk.

    14 May 2004: Test power supply

    I started construction of the first test power supply for the amplifier. The aim was to build an economical power supply using a toroidal transformer and some new 35 Volt electrolytic reservoir capacitors which I had on-hand. This will allow some familiarity to be gained with the LM4780 without stretching the absolute design limits and committing a lot of money to buying parts which may not be suitable in the long run. A 160 VA toroidal transformer (MT2114) was purchased from Jaycar for about AUD$50. It has two 25 volt secondary windings and is a "Made in China" no-name brand. A 400 Volt 10 Amp bridge rectifier (BR-104) was put to use with a pair of 15,000 uF 35 V Elna capacitors. I knew that the "off load" voltage would be more than 35 Volts, as these tests with dummy load resistors and a variac on the mains input show:

    Rload (Ohms)VprimaryV+ (aver)Vripple (p-p)
    Approx. 0.007 Amps,
    0.5 Watts total +/-
    250Not testedNot tested
    Approx. 0.2 Amps,
    14 Watts total +/-
    Approx. 3.5 Amps,
    200 Watts total +/-

    Photograph and Schematic Copyright Glenn Baddeley 2004
    The first test power supply as at 14 May 2004. The photograph shows a connected dummy load which consists of sixteen 1 Ohm 50 Watts resistors, configured as two 8 Ohm resistors rated at 400 Watts each, mounted on a hefty heatsink.

    27 May 2004: Extended primary on power supply transformer

    The solution to the excess voltage problem was to "extend" the primary winding of the transformer. A test winding of 10 turns measured 2.06 V under load, indicating that the transformer was designed at 5 turns per volt, giving 1200 turns on the 240 V primary and 125 turns on each of the 25 V secondary windings. Using a small shuttle, I wound 50 turns of two strands of 0.6mm enamelled wire on to the toroid, with a tap at 25 turns on one of the strands. This is a fairly fiddly and labourous process and I would not recommend it to the beginner electronics constructor. It took me nearly half an hour to carefully thread the 50 turns and to make sure they were neat. By using the two extra windings in series combinations with the existing primary, the 25 V nominal secondary could be reduced by up to 3 V in 0.5 V steps. Some tests with dummy load resistors and a variac show:

    Primary TurnsRload (Ohms)VprimaryV+ (aver)
    (= 240 + 5%)

    The choice was to go with 1275 primary turns (1200 turns on original primary + 50 turns on 1st extension + 25 turn tap on 2nd extension). This will mean that the 35 V rating of the capacitors will not be exceeded when the LM4780 is idling and the mains voltage is 5% high.

    Photograph and Schematic Copyright Glenn Baddeley 2004
    The power supply for the first prototype amplifier as at 27 May 2004. The photograph shows a dummy load connected which consists of two 180 Ohm 20 Watts resistors, which is more than the quiescent current of a LM4780. The foreground portion of the toroid has been over-wound with two windings of 50 turns each. One of them has a tap at 25 turns. This enables the transformer output to be decreased or increased, depending on how the additional windings are connected in series with the primary winding.

    1 June 2004: Prototype constructed on perforated board

    I've designed a schematic for a prototype stereo power amplifier and it has been assembled on a small piece of perforated board. The board has an array of holes separated by 0.1 inch and a copper pad for each hole. The hard-wiring technique can be a bit fiddly, but its faster than designing, etching and drilling a one-off PCB which will almost certainly require some changes. The prototype worked as designed. I will provide the test results and schematic in due course.

    Photographs Copyright Glenn Baddeley 2004

    28 June 2004: "NC" pins are not connected

    Email correspondence with the Design Support Group at National Semiconductor.

    I wrote on 24 June 2004 ' The connection diagram on page 2 shows pins 1, 3, 6, 13, 23, 24 and 26 as "NC", which normally represents "No Connection". The usage or limitation of these pins are not further discussed in the specifications. Inspection of the art work for the LM4780 Reference PCB on page 22 shows that:
    * pins 1 and 3 are connected to +Vcc (probably because adjacent pin 5 is a +Vcc pin??)
    * pin 6 is not connected to anything
    * pins 13, 23 and 24 are connected to GND (pins 12 and 19).
    My question is to confirm the status and purpose of the "NC" pins.
    1. Are any of the NC pins connected to any part of the internal circuitry ? (ie. the NC pins physically enter the package, but are they connected internally to the silicon dice or to any other pins)
    2. Are there any requirements for any of the NC pins to be externally connected to other pins? (such as in the Reference PCB mentioned above)

    The response on 26 June 2004 was 'The "NC" pins are not connected to the chip internally and it is permissable to tie them to an external signal without causing problems with the part. "NC" means "no connect" and the pins are not connected internally'.

    That's good news for the PCB designer, and I'm going to take advantage of that fact in my first PCB design for an LM4780 stereo power amplifier.

    26 August 2004: Role of the "GND" pins

    More email correspondence with the Design Support Group.

    I wrote on 24 August 2004 'The schematics of the Reference / Evaluation Boards in the LM4780 / 81 / 82 PDF data sheets do not quite match the PCB artwork and the Bill Of Materials. There is an additional mysterious resistor RG of 2.7 Ohms on the PCB which connects the GND pins to the 0V centre of the power supply. The GND pins are also part of a "star earthing" system which involves the input stage (Rin1, Rin2, Cin1, Cin2 & Input sockets ), the feedback network (Ci1, Ci2) and the mute switch (S1). Is this the same as a technique which is commonly used in discrete audio power amplifier design to isolate the input stage currents from the output stage currents, and to reduce "earth loop" effects? I have noted that the GND pins sink a current of about 6.5 ma towards the -Vee rail, and this effectively increases the DC offest voltage on the output by 17.5 mV (= 2.7 * 0.0065 ). This being the case, it would not be recommended to increase RG much above 3 Ohms, as the output DC current flowing through the loudspeaker voice coil would be above the commonly accepted limit of 20 mV. None of the example schematics use RG. Is the performance of the amplifier compromised in any respect if RG were replaced with a 0 Ohm link on the Reference board?'.

    The response on 26 August 2004 was 'The purpose of the Rg resistor is to create a small amount of isolation between the 'dirty' high current GNDs and the very low current 'clean' GNDs. The resistor can be safely shorted or changed in value. The only limitation is that the value stay very low, as you have already noted. The graphs in the LM4780/81/82 were all created using the demo boards that are available...'.

    27 August 2004: LM3886 equivalent schematic details

    More email correspondence with the Design Support Group regarding a couple of "loose ends" in the Equivalent Schematic for the LM3886.

    I wrote on 24 August 2004 something similar to 'Page 6 of the October 2003 PDF data sheet for the LM3886 contains an Equivalent Schematic which has a 3 'triangular' terminals, none of which is marked as GND, and 2 other 't bar' terminals on the collectors of 2 transistors in the lower half of the schematic, and one other inverted 't bar' terminal in the upper half of the schematic, none of which are marked as connecting to any other part of the schematic'.

    The response on 27 August 2004 was 'The upper 'T' connection goes to V- eventually through a current source. There is more to the circuit than shown but it is really just a current source for biasing. The bottom 'T' connections go to the same line which goes to the emitter of a transistor that has it's collector connected to V+. So they are connected to some voltage below V+ but above GND (0V). There are 3 unfilled triangles in the LM3886 equivalent circuit and these all indicate a connection to GND.'

    I have produced an annotated equivalent schematic of the LM3886 which shows this arrangement. The complete schematic is confidential and will probably never be released to the public by National Semiconductor.

    24 September 2004: LM4780 is a dual LM3886

    Refer to the LM4780 / 3886 / 3875 / 1875 Comparison page which contains electrical characteristic tables, equivalent schematics and distortion graphs for these four chips. Based purely on the electrical characteristics, the LM4780 appears to be dual version of the LM3886. The following parameters have exactly the same values:

    The LM4780 has a Signal-to-Noise ratio (Po=1 W, A-Weighted, Measured at 1 KHz, Rs=25 Ohm) of 97 dB, which is slightly better than the LM3886 Signal-to-Noise ratio of 92.5 dB.

    With +/- 28 Volt supply rails the LM3886 is specified to deliver 68 Watts into 4 Ohms, at 0.1% THD+N and 20 Hz - 20 KHz bandwidth. The LM4780 is only specified at +/-25 Volts rails into 4 Ohms, and delivers 55 Watts, at 1% THD. However, according to the graph at middle left on page 10 of the PDF spec sheet, the power supply rails can be lifted to +/-28 Volts, where upon it will deliver approximately 66 Watts. I don't think LM4780 is any less capable than the LM3886 when driving 4 Ohm loads (eg. they have the same peak current output capability), its just that the sample performance specifications are expressed differently.

    The LM3886 is specified at +/-28 V (for 4 Ohm and 8 Ohm loads) and +/-35 V (for 8 Ohm), whereas the LM4780 is specified at +/-25 V (4 Ohm), +/-30V (6 Ohm) and +/-35V (8 Ohm).

    I had more email correspondence with the Design Support Group on 24 August 2004 where I wrote 'The LM3886 has almost identical characteristics to half of an LM4780, so I assume their design is very similar'.

    The response on 26 August 2004 was 'The LM4780 is two LM3886 die in a single package. So the equivalent schematic from the LM3886 DS can be used as an equivalent schematic for each channel (amplifier) in the LM4780. With the LM4780 being two die the GND pins are NOT internally connected'. A further response on 27 August 2004 was 'There is no difference in the two die used in the LM4780 and the single die used in the LM3886. Using two die is more efficient thermally as the distance between the die helps improve heat flow out of the package to the heat sink. Cost is also reduced as smaller die have higher yields than larger die due to defect density so a single, large dual die design would not have any benefits over two separate die'.

    I wrote again on 30 August 2004 'From the connection diagram on page 2 of the LM4780 PDF data sheet it is fairly clear how the lead-outs from the two LM3886 dice are allocated to pins 7 & 12-16 for die "A" and pins 19-25 for die "B", but I am confused about the five pairs of V+ and V- pins. Are the five V+ pins connected together internally, and the five V- pins connected together internally, or are they individually distributed between the two dice?'.

    The Design Support Group responsed on 14 September 2004 'Here is the connection of each die:
    Pin Connection
    1 N/C
    2 V- (Vee) Down bond to the DAP of the package
    3 N/C
    4 V- (Vee) to die A 1st V- pad
    5 V+ (Vcc) to die A 1st V+pad
    6 N/C
    7 Output from die A pad
    8 V+ (Vcc) to die A 2nd V+ pad
    9 V+ (Vcc) to die A 2nd V+ pad
    10 V- (Vee) to die A 2nd V- pad
    11 V- (Vee) to die A 2nd V- pad
    12 GND to die A GND pad
    13 N/C
    14 Mute to die A Mute pad
    15 INA- to die A INA- pad
    16 INA+ to die A INA+ pad
    17 V+ (Vcc) to die B 2nd V+ pad
    18 V- (Vee) to die B 2nd V- pad
    19 GND to die B GND pad
    20 Mute to die B Mute pad
    21 INB- to die B INB- pad
    22 INB+ to die B INB+ pad
    23 N/C
    24 N/C
    25 Output from die B pad
    26 N/C
    27 V+ (Vcc) to die B 1st V+pad
    Internal Downbond to DAP from die B 1st V- pad

    The reason sometimes there are two V- or V+ pins to a pad is that there is one bond wire from each pin. This makes for better connection to the PCB. When there is only one pin then there are two bond wires from the pin to the pad. Downbonds are all double wires as well. Inputs, Mute and GND are all single bond wires since they are low current'.

    After I asked for some clarification on a couple of the connections, there was a further response on 15 September 2004 'DAP means Die Attach Pad, it is the flat area where the die is attached to the package. .... There is only one wire from pin 4 to the V- pad because of spacing it was not possible to do two wires. Also, the output pins have two wires to them but each die actually has two output pads. One wire from each pad goes to pin 7 or pin 25 respectively. So two wires to each output pin'.

    I have summarised this information concerning the LM4780 internal connection scheme to the two LM3886 die in two annotated pin connection diagrams on the LM4780 Dual 60W Audio Power Amplifier page.

    I have also added my version of an annotated equivalent schematic for the LM4780 on the LM4780 / 3886 / 3875 / 1875 Comparison page.

    The slightly different Distortion versus Power Output characteristics between the LM4780 and LM3886 would be mostly due to the different packaging and layout of the Reference PCB. The characteristic curves are measured and plotted by National Semiconductor with the devices installed on their Reference PCBs.

    17 January 2005: Prototype demonstrated to Melbourne Audio Club

    I demonstrated my prototype LM4780 stereo power amplifier on the 15th December 2004 at the General Meeting of the Melbourne Audio Club, which is the annual "DIY night". A report on the meeting was published in "Melbourne Audio News", Issue 353, January 2005. The MAC Program Co-ordinator, Bob Toll, wrote:

    Next Glenn Baddeley presented a complete system with a TEAC CD player modified with a two valve output section, a chip amp he constructed using the LM4780, and two way floor standing speakers using Focal and Dynaudio drivers. The cabinets were completely inert being constructed using five different layers including lead, Masonite and MDF; and the drivers were time-aligned. Great audio.

    Another LM4780 amp was demonstrated at the same meeting with a custom PCB design based on the National Semiconductor evaluation design. Bob wrote:

    Chris Mogford presented a similar amplifier with a computer style Creative CD player (which refused to select tracks so the TEAC from the previous system was commandeered) with a pair of largish book-shelf speakers built around German Visaton and Jaycar drivers by Bob Toll, with crossovers by Ron Newbound. Ron also supplied a small powere subwoofer to augment the bass. Very detailed.

    Photograph Copyright Bob Field 2004
    Glenn Baddeley removing the cover from a power supply. It provides the low voltage and high voltage requirements of a modified TEAC CD player which has a valve output stage. The CD player is on the left and one of the Focal / Dynaudio speakers is on the extreme right.

    Photograph Copyright Glenn Baddeley 2004
    A view of three audio systems set up on stage in the Willis Room at the Nunawading Civic Centre.

    Photograph Copyright Glenn Baddeley 2004
    A rear view of Glenn's LM4780 prototype power amplifier. There is no case, everything is mounted on square section aluminium tubing which is attached to the heat sink.

    Photographs Copyright Glenn Baddeley 2005
    Two more views of Glenn's LM4780 prototype power amplifier (on the workbench at home).

    Schematics for the prototype power amplifer, including the power supply. All resistors are 0.25 W, 1 % metal film, except the 4.7 Ohm resistors which are 1W, 5% carbon. All non-electrolytic capacitors are 100 V, 5% metallised polystyrene.

    Photographs Copyright Bob Field 2004
    Chris Mogford and his LM4780 stereo amplifier PCB. PDF files are available for the PCB design (61K) and the schematic (20K).

    [ Top ]   [ Back to LM4780 Amplifier Page ]

    © Glenn Baddeley 2005.
    http://home.pacific.net.au/~gnb/audio/lm4780designlog.html was last updated 10 February 2005.
    Report problems and send comments to Glenn Baddeley.