Best way to plot log-frequency spectrograms?

[danferns]

Hello! Thank you for making an incredibly fast plotting library.

I want to use xy to plot large STFTs. I'm able to get a decent plot using the heatmap chart:

    chart = xy.heatmap_chart(xy.heatmap(data, x=times, y=freqs, colormap="viridis"))
    chart.write_image("test.png")

However, it is more useful to interpret a spectrogram with the frequency axis (y-axis) having a logarithmic scale. But when I try the following, it loses resolution along the frequency axis. What gives?

    freq_axis = xy.y_axis(
            id="yfreq",
            label="freq",
            type_="log",
            domain=(20, sr/2)
    )
    chart = xy.heatmap_chart(xy.heatmap(data, x=times, y=freqs, colormap="viridis", y_axis="yfreq"), freq_axis)
    chart.write_image("test_log.png")

The matplotlib API gives me a result closer to what I'm aiming for. This is a port from the code I used to have:

    extent = [times[0], times[-1], freqs[0], freqs[-1]]
    ax1.imshow(
        data,
        extent=extent,
        aspect="auto",
        origin="lower",
        cmap="viridis",
        interpolation="bilinear",
    )
    ax1.set_yscale("log")
    ax1.set_ylim(20, sr / 2)
    ax1.grid(False)

    plt.show()

Notice we have more detail as we go higher up the image.

However, this only looks good with plt.show() (aka. it opens in the browser). plt.savefig() does not have great resolution. And both these methods appear to take longer than using the chart API.

What would be the recommended approach for my use case?