Accent level in 1.023

Discuss use it and firmware update
  • +1

    The difference between BD without accent and with accent is almost not notable.
  • There are strange connection between attack and decay also. it is connected with volume. So global settings have to be fucked!
  • Getting closer to the schematics, I notice a major difference between the TR909 and the Nava... Vincent, please correct me if I am wrong.

    In the original TR909 design, the sequencer drives 4174 6-bit D-latch registers, which are used as DAC with series of resistors. The 6-bit value issued of each 4171 is latched for the duration of each step. If the instrument is not triggered, the accent control signal keeps the value of the last step in which the instrument was played.

    The CD4051 is not a latching demultiplexer. The value is driven on each step by the sequencer. However the value is not maintained. A capacitor between the output of the 4051 and the non-inverting inputof the accent op-amp (i.e. C89 for the Bassdrum) produces a decay on the accent signal. Moreover, if the instrument is not triggered on a step, the accent takes the maximum value instead of the previous value.

    The rest of the analog design is the same.

    The snapshot below shows the signal driving the accent of the bassdrum of the TR909 in red and of the Nava in blue (DC coupling).
    Image

    The snapshot below shows the ENV1 of the bassdrum of the TR909 in red and of the Nava in blue (DC coupling).
    Image

    The snapshot below shows the accent and trigger signals of the BD of the Nava, respectively in red and blue.
    Image

    The pattern shown in the waveform is as follows: 1:ON+ACC 2:OFF 3:ON 4:OFF 5:ON+ACC 6:OFF 7:ON 8:OFF 9:ON+ACC 10:OFF 11:ON 12:OFF 13:ON+ACC 14:OFF 15:ON 16:OFF

    To my analysis, there are bugs:

    1. My primary question is about the use of CD4051 as non-latched demultiplexors:
    Assuming that the use of the CD4051 with non-latched outputs was intentional, the select bits (MUX_A,MUX_B,MUX_C,INH_MUX1,INH_MUX2) shall actually coutinuously round to keep the value stable using the input capacitors of the accent circuits. However looking at the waveforms of these signals, this is not the behavior of the software. It appears that the outputs of each CD4051 are selected only once per step.

    2. The peak level of the accent of the Nava does not match the level of the accent of the TR909.

    3. The accent level shall be maintained on steps where the instrument is not played
  • The CD4051 demultiplexors could be replaced by Analog Device SMP18FPZ but they are end of life and sold at 17€ each on mouser.
    Another option could be to replace the CD4051 by daughter boards adding sample and hold circuits on the 8 outputs of the CD4051 but that requires to design and build mini-pcbs...
  • There are SN74LV4051A or CD74HC4051 and DG9051 even..
  • lukep_gdansk wrote:There are SN74LV4051A or CD74HC4051 and DG9051 even..
    These won't work. The outputs must be held when not selected by the (A,B,C,INH) bus, resp. (MUX_A,MUX_B,MUX_C,INH_MUX1/2) in the sequencer schematic. The issue is the leakage which is too high to keep the value long enough.

    So either a daughter board with a DIP16 male connector must be build to add sample&hold circuits to the outputs of the CD4051, or the CD4051 must be substituted by a 3-to-8 demultiplexor with embedded sample&hold. Note that in the latter case, we have to be cautious about the required sample and the hold time.

    Another solution could be to create a daughter board with a DIP16 male connector to replace the CD4051, which would reuse the (MUX_A,MUX_B,MUX_C,INH_MUX1/2) control signals from the processor to drive shift registers. The parallel outputs of the shift registers would then drive R/2R ladder resistor networks to make the dac convertion. But the required area might be important. In that case, the DAC would not be used to control the accent level. The 6-bit output of the shift registers would then drive resistor networks as in the original design. This guarantees the stability of the accent voltage throughout the duration of the step.

    Any of these options would have an impact of the software, which means that two versions should then be maintained: one for the original design, one with the updated design, whether the solution is to use sample & hold circuits or shift registers.

    For that last solution, 74HC595N can be used as shift-registers (chained). DAC conversion made using R/2R networks such as Bourns 4610X-R2R-103LF.
    Each CD4051 drives 5 accent signals, with a 6-bit resolution each, so each daughter board plugged as replacement of CD4051 could be of 5 cascaded 74HC595N where RCK <= INH_MUX1/2; SER <= MUX_A; SCK <= MUX_B/C (to be confirmed). The 8-bit output of each 74HC595N would then drive an R/2R network as DAC. The DAC resolution would then become 8-bit. Though the resolution is 6-bit in the original TR-909, this would allow using the 7-bit resolution of the MIDI velocity.