It is difficult to create appropriate bar charts for data that cover large value ranges. The usual approach for these cases employs a logarithmic scale, which, however, suffers from issues inherent to its non-linear mapping: for example, a quantitative comparison of different values is difficult. We present a new approach for bar charts that combines the advantages of linear and logarithmic scales, while avoiding their drawbacks. Our scale-stack bar charts use multiple scales to cover a large value range, while the linear mapping within each scale preserves the ability to visually compare quantitative ratios. Scale-stack bar charts can be used for the same applications as classic bar charts; in particular, they can readily handle stacked bar representations and negative values. Our visualization technique is demonstrated with results for three different application areas and is assessed by an expert review and a quantitative user study confirming advantages of our technique for quantitative comparisons.